Fix variable quoting and profile path in optimized_clang.sh

Co-authored-by: mykaul <4655593+mykaul@users.noreply.github.com>

.github/CODEOWNERS

@@ -1,5 +1,5 @@
 # AUTH
-auth/* @nuivall @ptrsmrn
+auth/* @nuivall
 
 # CACHE
 row_cache* @tgrabiec
@@ -25,11 +25,11 @@ compaction/* @raphaelsc
 transport/*
 
 # CQL QUERY LANGUAGE
-cql3/* @tgrabiec @nuivall @ptrsmrn
+cql3/* @tgrabiec @nuivall
 
 # COUNTERS
-counters* @nuivall @ptrsmrn
-tests/counter_test* @nuivall @ptrsmrn
+counters* @nuivall
+tests/counter_test* @nuivall
 
 # DOCS
 docs/* @annastuchlik @tzach

.github/scripts/auto-backport.py

@@ -62,7 +62,7 @@ def create_pull_request(repo, new_branch_name, base_branch_name, pr, backport_pr
         if is_draft:
             labels_to_add.append("conflicts")
             pr_comment = f"@{pr.user.login} - This PR was marked as draft because it has conflicts\n"
-            pr_comment += "Please resolve them and mark this PR as ready for review"
+            pr_comment += "Please resolve them and remove the 'conflicts' label. The PR will be made ready for review automatically."
             backport_pr.create_issue_comment(pr_comment)
         
         # Apply all labels at once if we have any

.github/workflows/backport-pr-fixes-validation.yaml

@@ -18,7 +18,7 @@ jobs:
             
             // Regular expression pattern to check for "Fixes" prefix
             // Adjusted to dynamically insert the repository full name
-            const pattern = `Fixes:? (?:#|${repo.replace('/', '\\/')}#|https://github\\.com/${repo.replace('/', '\\/')}/issues/)(\\d+)`;
+            const pattern = `Fixes:? ((?:#|${repo.replace('/', '\\/')}#|https://github\\.com/${repo.replace('/', '\\/')}/issues/)(\\d+)|([A-Z]+-\\d+))`;
             const regex = new RegExp(pattern);
             
             if (!regex.test(body)) {

.github/workflows/trigger-scylla-ci.yaml

@@ -3,10 +3,13 @@ name: Trigger Scylla CI Route
 on:
   issue_comment:
     types: [created]
+  pull_request_target:
+    types:
+      - unlabeled
 
 jobs:
   trigger-jenkins:
-    if: github.event.comment.user.login != 'scylladbbot' && contains(github.event.comment.body, '@scylladbbot') && contains(github.event.comment.body, 'trigger-ci')
+    if: (github.event.comment.user.login != 'scylladbbot' && contains(github.event.comment.body, '@scylladbbot') && contains(github.event.comment.body, 'trigger-ci')) || github.event.label.name == 'conflicts'
     runs-on: ubuntu-latest
     steps:
       - name: Trigger Scylla-CI-Route Jenkins Job

alternator/conditions.cc

@@ -42,7 +42,7 @@ comparison_operator_type get_comparison_operator(const rjson::value& comparison_
     if (!comparison_operator.IsString()) {
         throw api_error::validation(fmt::format("Invalid comparison operator definition {}", rjson::print(comparison_operator)));
     }
-    std::string op = comparison_operator.GetString();
+    std::string op = rjson::to_string(comparison_operator);
     auto it = ops.find(op);
     if (it == ops.end()) {
         throw api_error::validation(fmt::format("Unsupported comparison operator {}", op));
@@ -377,8 +377,8 @@ bool check_compare(const rjson::value* v1, const rjson::value& v2, const Compara
         return cmp(unwrap_number(*v1, cmp.diagnostic), unwrap_number(v2, cmp.diagnostic));
     }
     if (kv1.name == "S") {
-        return cmp(std::string_view(kv1.value.GetString(), kv1.value.GetStringLength()),
-                   std::string_view(kv2.value.GetString(), kv2.value.GetStringLength()));
+        return cmp(rjson::to_string_view(kv1.value),
+                   rjson::to_string_view(kv2.value));
     }
     if (kv1.name == "B") {
         auto d_kv1 = unwrap_bytes(kv1.value, v1_from_query);
@@ -470,9 +470,9 @@ static bool check_BETWEEN(const rjson::value* v, const rjson::value& lb, const r
         return check_BETWEEN(unwrap_number(*v, diag), unwrap_number(lb, diag), unwrap_number(ub, diag), bounds_from_query);
     }
     if (kv_v.name == "S") {
-        return check_BETWEEN(std::string_view(kv_v.value.GetString(), kv_v.value.GetStringLength()),
-                             std::string_view(kv_lb.value.GetString(), kv_lb.value.GetStringLength()),
-                             std::string_view(kv_ub.value.GetString(), kv_ub.value.GetStringLength()),
+        return check_BETWEEN(rjson::to_string_view(kv_v.value),
+                             rjson::to_string_view(kv_lb.value),
+                             rjson::to_string_view(kv_ub.value),
                              bounds_from_query);
     }
     if (kv_v.name == "B") {

alternator/consumed_capacity.cc

@@ -8,6 +8,8 @@
 
 #include "consumed_capacity.hh"
 #include "error.hh"
+#include "utils/rjson.hh"
+#include <fmt/format.h>
 
 namespace alternator {
 
@@ -32,12 +34,12 @@ bool consumed_capacity_counter::should_add_capacity(const rjson::value& request)
     if (!return_consumed->IsString()) {
         throw api_error::validation("Non-string ReturnConsumedCapacity field in request");
     }
-    std::string consumed = return_consumed->GetString();
+    std::string_view consumed = rjson::to_string_view(*return_consumed);
     if (consumed == "INDEXES") {
         throw api_error::validation("INDEXES consumed capacity is not supported");
     }
     if (consumed != "TOTAL") {
-        throw api_error::validation("Unknown consumed capacity "+ consumed);
+        throw api_error::validation(fmt::format("Unknown consumed capacity {}", consumed));
     }
     return true;
 }

alternator/executor.cc

@@ -419,7 +419,7 @@ static std::optional<std::string> find_table_name(const rjson::value& request) {
     if (!table_name_value->IsString()) {
         throw api_error::validation("Non-string TableName field in request");
     }
-    std::string table_name = table_name_value->GetString();
+    std::string table_name = rjson::to_string(*table_name_value);
     return table_name;
 }
 
@@ -546,7 +546,7 @@ get_table_or_view(service::storage_proxy& proxy, const rjson::value& request) {
             // does exist but the index does not (ValidationException).
             if (proxy.data_dictionary().has_schema(keyspace_name, orig_table_name)) {
                 throw api_error::validation(
-                    fmt::format("Requested resource not found: Index '{}' for table '{}'", index_name->GetString(), orig_table_name));
+                    fmt::format("Requested resource not found: Index '{}' for table '{}'", rjson::to_string_view(*index_name), orig_table_name));
             } else {
                 throw api_error::resource_not_found(
                     fmt::format("Requested resource not found: Table: {} not found", orig_table_name));
@@ -587,7 +587,7 @@ static std::string get_string_attribute(const rjson::value& value, std::string_v
         throw api_error::validation(fmt::format("Expected string value for attribute {}, got: {}",
                 attribute_name, value));
     }
-    return std::string(attribute_value->GetString(), attribute_value->GetStringLength());
+    return rjson::to_string(*attribute_value);
 }
 
 // Convenience function for getting the value of a boolean attribute, or a
@@ -1080,8 +1080,8 @@ static void add_column(schema_builder& builder, const std::string& name, const r
     }
     for (auto it = attribute_definitions.Begin(); it != attribute_definitions.End(); ++it) {
         const rjson::value& attribute_info = *it;
-        if (attribute_info["AttributeName"].GetString() == name) {
-            auto type = attribute_info["AttributeType"].GetString();
+        if (rjson::to_string_view(attribute_info["AttributeName"]) == name) {
+            std::string_view type = rjson::to_string_view(attribute_info["AttributeType"]);
             data_type dt = parse_key_type(type);
             if (computed_column) {
                 // Computed column for GSI (doesn't choose a real column as-is
@@ -1116,7 +1116,7 @@ static std::pair<std::string, std::string> parse_key_schema(const rjson::value&
         throw api_error::validation("First element of KeySchema must be an object");
     }
     const rjson::value *v = rjson::find((*key_schema)[0], "KeyType");
-    if (!v || !v->IsString() || v->GetString() != std::string("HASH")) {
+    if (!v || !v->IsString() || rjson::to_string_view(*v) != "HASH") {
         throw api_error::validation("First key in KeySchema must be a HASH key");
     }
     v = rjson::find((*key_schema)[0], "AttributeName");
@@ -1124,14 +1124,14 @@ static std::pair<std::string, std::string> parse_key_schema(const rjson::value&
         throw api_error::validation("First key in KeySchema must have string AttributeName");
     }
     validate_attr_name_length(supplementary_context, v->GetStringLength(), true, "HASH key in KeySchema - ");
-    std::string hash_key = v->GetString();
+    std::string hash_key = rjson::to_string(*v);
     std::string range_key;
     if (key_schema->Size() == 2) {
         if (!(*key_schema)[1].IsObject()) {
             throw api_error::validation("Second element of KeySchema must be an object");
         }
         v = rjson::find((*key_schema)[1], "KeyType");
-        if (!v || !v->IsString() || v->GetString() != std::string("RANGE")) {
+        if (!v || !v->IsString() || rjson::to_string_view(*v) != "RANGE") {
             throw api_error::validation("Second key in KeySchema must be a RANGE key");
         }
         v = rjson::find((*key_schema)[1], "AttributeName");
@@ -1887,8 +1887,8 @@ future<executor::request_return_type> executor::create_table(client_state& clien
         std::string def_type = type_to_string(def.type);
         for (auto it = attribute_definitions.Begin(); it != attribute_definitions.End(); ++it) {
             const rjson::value& attribute_info = *it;
-            if (attribute_info["AttributeName"].GetString() == def.name_as_text()) {
-                auto type = attribute_info["AttributeType"].GetString();
+            if (rjson::to_string_view(attribute_info["AttributeName"]) == def.name_as_text()) {
+                std::string_view type = rjson::to_string_view(attribute_info["AttributeType"]);
                 if (type != def_type) {
                     throw api_error::validation(fmt::format("AttributeDefinitions redefined {} to {} already a key attribute of type {} in this table", def.name_as_text(), type, def_type));
                 }
@@ -2362,7 +2362,7 @@ put_or_delete_item::put_or_delete_item(const rjson::value& item, schema_ptr sche
     _cells = std::vector<cell>();
     _cells->reserve(item.MemberCount());
     for (auto it = item.MemberBegin(); it != item.MemberEnd(); ++it) {
-        bytes column_name = to_bytes(it->name.GetString());
+        bytes column_name = to_bytes(rjson::to_string_view(it->name));
         validate_value(it->value, "PutItem");
         const column_definition* cdef = find_attribute(*schema, column_name);
         validate_attr_name_length("", column_name.size(), cdef && cdef->is_primary_key());
@@ -2783,10 +2783,10 @@ static void verify_all_are_used(const rjson::value* field,
         return;
     }
     for (auto it = field->MemberBegin(); it != field->MemberEnd(); ++it) {
-        if (!used.contains(it->name.GetString())) {
+        if (!used.contains(rjson::to_string(it->name))) {
             throw api_error::validation(
                 format("{} has spurious '{}', not used in {}",
-                    field_name, it->name.GetString(), operation));
+                    field_name, rjson::to_string_view(it->name), operation));
         }
     }
 }
@@ -3000,7 +3000,7 @@ future<executor::request_return_type> executor::delete_item(client_state& client
 }
 
 static schema_ptr get_table_from_batch_request(const service::storage_proxy& proxy, const rjson::value::ConstMemberIterator& batch_request) {
-    sstring table_name = batch_request->name.GetString(); // JSON keys are always strings
+    sstring table_name = rjson::to_sstring(batch_request->name); // JSON keys are always strings
     try {
         return proxy.data_dictionary().find_schema(sstring(executor::KEYSPACE_NAME_PREFIX) + table_name, table_name);
     } catch(data_dictionary::no_such_column_family&) {
@@ -3055,17 +3055,44 @@ public:
     }
 };
 
-static future<> cas_write(service::storage_proxy& proxy, schema_ptr schema, service::cas_shard cas_shard, const dht::decorated_key& dk, const std::vector<put_or_delete_item>& mutation_builders,
-        service::client_state& client_state, tracing::trace_state_ptr trace_state, service_permit permit) {
+future<> executor::cas_write(schema_ptr schema, service::cas_shard cas_shard, const dht::decorated_key& dk,
+        const std::vector<put_or_delete_item>& mutation_builders, service::client_state& client_state,
+        tracing::trace_state_ptr trace_state, service_permit permit)
+{
+    if (!cas_shard.this_shard()) {
+        _stats.shard_bounce_for_lwt++;
+        return container().invoke_on(cas_shard.shard(), _ssg,
+                    [cs = client_state.move_to_other_shard(),
+                    &mb = mutation_builders,
+                    &dk,
+                    ks = schema->ks_name(),
+                    cf = schema->cf_name(),
+                    gt = tracing::global_trace_state_ptr(trace_state),
+                    permit = std::move(permit)]
+                    (executor& self) mutable {
+            return do_with(cs.get(), [&mb, &dk, ks = std::move(ks), cf = std::move(cf),
+                                    trace_state = tracing::trace_state_ptr(gt), &self]
+                                    (service::client_state& client_state) mutable {
+                auto schema = self._proxy.data_dictionary().find_schema(ks, cf);
+                service::cas_shard cas_shard(*schema, dk.token());
+
+                //FIXME: Instead of passing empty_service_permit() to the background operation,
+                // the current permit's lifetime should be prolonged, so that it's destructed
+                // only after all background operations are finished as well.
+                return self.cas_write(schema, std::move(cas_shard), dk, mb, client_state, std::move(trace_state), empty_service_permit());
+            });
+        });
+    }
+
     auto timeout = executor::default_timeout();
     auto op = std::make_unique<put_or_delete_item_cas_request>(schema, mutation_builders);
     auto* op_ptr = op.get();
     auto cdc_opts = cdc::per_request_options{
         .alternator = true,
         .alternator_streams_increased_compatibility =
-                schema->cdc_options().enabled() && proxy.data_dictionary().get_config().alternator_streams_increased_compatibility(),
+                schema->cdc_options().enabled() && _proxy.data_dictionary().get_config().alternator_streams_increased_compatibility(),
     };
-    return proxy.cas(schema, std::move(cas_shard), *op_ptr, nullptr, to_partition_ranges(dk),
+    return _proxy.cas(schema, std::move(cas_shard), *op_ptr, nullptr, to_partition_ranges(dk),
             {timeout, std::move(permit), client_state, trace_state},
             db::consistency_level::LOCAL_SERIAL, db::consistency_level::LOCAL_QUORUM,
             timeout, timeout, true, std::move(cdc_opts)).finally([op = std::move(op)]{}).discard_result();
@@ -3091,13 +3118,11 @@ struct schema_decorated_key_equal {
 
 // FIXME: if we failed writing some of the mutations, need to return a list
 // of these failed mutations rather than fail the whole write (issue #5650).
-static future<> do_batch_write(service::storage_proxy& proxy,
-        smp_service_group ssg,
+future<> executor::do_batch_write(
         std::vector<std::pair<schema_ptr, put_or_delete_item>> mutation_builders,
         service::client_state& client_state,
         tracing::trace_state_ptr trace_state,
-        service_permit permit,
-        stats& stats) {
+        service_permit permit) {
     if (mutation_builders.empty()) {
         return make_ready_future<>();
     }
@@ -3119,7 +3144,7 @@ static future<> do_batch_write(service::storage_proxy& proxy,
             mutations.push_back(b.second.build(b.first, now));
             any_cdc_enabled |= b.first->cdc_options().enabled();
         }
-        return proxy.mutate(std::move(mutations),
+        return _proxy.mutate(std::move(mutations),
                 db::consistency_level::LOCAL_QUORUM,
                 executor::default_timeout(),
                 trace_state,
@@ -3128,7 +3153,7 @@ static future<> do_batch_write(service::storage_proxy& proxy,
                 false,
                 cdc::per_request_options{
                     .alternator = true,
-                    .alternator_streams_increased_compatibility = any_cdc_enabled && proxy.data_dictionary().get_config().alternator_streams_increased_compatibility(),
+                    .alternator_streams_increased_compatibility = any_cdc_enabled && _proxy.data_dictionary().get_config().alternator_streams_increased_compatibility(),
                 });
     } else {
         // Do the write via LWT:
@@ -3140,46 +3165,35 @@ static future<> do_batch_write(service::storage_proxy& proxy,
             schema_decorated_key_hash, 
             schema_decorated_key_equal>;
         auto key_builders = std::make_unique<map_type>(1, schema_decorated_key_hash{}, schema_decorated_key_equal{});
-        for (auto& b : mutation_builders) {
-            auto dk = dht::decorate_key(*b.first, b.second.pk());
-            auto [it, added] = key_builders->try_emplace(schema_decorated_key{b.first, dk});
+        for (auto&& b : std::move(mutation_builders)) {
+            auto [it, added] = key_builders->try_emplace(schema_decorated_key {
+                .schema = b.first,
+                .dk = dht::decorate_key(*b.first, b.second.pk())
+            });
             it->second.push_back(std::move(b.second));
         }
         auto* key_builders_ptr = key_builders.get();
-        return parallel_for_each(*key_builders_ptr, [&proxy, &client_state, &stats, trace_state, ssg, permit = std::move(permit)] (const auto& e) {
-            stats.write_using_lwt++;
+        return parallel_for_each(*key_builders_ptr, [this, &client_state, trace_state, permit = std::move(permit)] (const auto& e) {
+            _stats.write_using_lwt++;
             auto desired_shard = service::cas_shard(*e.first.schema, e.first.dk.token());
-            if (desired_shard.this_shard()) {
-                return cas_write(proxy, e.first.schema, std::move(desired_shard), e.first.dk, e.second, client_state, trace_state, permit);
-            } else {
-                stats.shard_bounce_for_lwt++;
-                return proxy.container().invoke_on(desired_shard.shard(), ssg,
-                            [cs = client_state.move_to_other_shard(),
-                             &mb = e.second,
-                             &dk = e.first.dk,
-                             ks = e.first.schema->ks_name(),
-                             cf = e.first.schema->cf_name(),
-                             gt =  tracing::global_trace_state_ptr(trace_state),
-                             permit = std::move(permit)]
-                            (service::storage_proxy& proxy) mutable {
-                    return do_with(cs.get(), [&proxy, &mb, &dk, ks = std::move(ks), cf = std::move(cf),
-                                              trace_state = tracing::trace_state_ptr(gt)]
-                                              (service::client_state& client_state) mutable {
-                        auto schema = proxy.data_dictionary().find_schema(ks, cf);
+            auto s = e.first.schema;
 
-                        // The desired_shard on the original shard remains alive for the duration
-                        // of cas_write on this shard and prevents any tablet operations.
-                        // However, we need a local instance of cas_shard on this shard
-                        // to pass it to sp::cas, so we just create a new one.
-                        service::cas_shard cas_shard(*schema, dk.token());
-
-                        //FIXME: Instead of passing empty_service_permit() to the background operation,
-                        // the current permit's lifetime should be prolonged, so that it's destructed
-                        // only after all background operations are finished as well.
-                        return cas_write(proxy, schema, std::move(cas_shard), dk, mb, client_state, std::move(trace_state), empty_service_permit());
-                    });
-                }).finally([desired_shard = std::move(desired_shard)]{});
-            }
+            static const auto* injection_name = "alternator_executor_batch_write_wait";
+            return utils::get_local_injector().inject(injection_name, [s = std::move(s)] (auto& handler) -> future<> {
+                const auto ks = handler.get("keyspace");
+                const auto cf = handler.get("table");
+                const auto shard = std::atoll(handler.get("shard")->data());
+                if (ks == s->ks_name() && cf == s->cf_name() && shard == this_shard_id()) {
+                    elogger.info("{}: hit", injection_name);
+                    co_await handler.wait_for_message(std::chrono::steady_clock::now() + std::chrono::minutes{5});
+                    elogger.info("{}: continue", injection_name);
+                }
+            }).then([&e, desired_shard = std::move(desired_shard),
+                 &client_state, trace_state = std::move(trace_state), permit = std::move(permit), this]() mutable
+            {
+                return cas_write(e.first.schema, std::move(desired_shard), e.first.dk,
+                    std::move(e.second), client_state, std::move(trace_state), std::move(permit));
+            });
         }).finally([key_builders = std::move(key_builders)]{});
     }
 }
@@ -3327,7 +3341,7 @@ future<executor::request_return_type> executor::batch_write_item(client_state& c
     _stats.wcu_total[stats::DELETE_ITEM] += wcu_delete_units;
     _stats.api_operations.batch_write_item_batch_total += total_items;
     _stats.api_operations.batch_write_item_histogram.add(total_items);
-    co_await do_batch_write(_proxy, _ssg, std::move(mutation_builders), client_state, trace_state, std::move(permit), _stats);
+    co_await do_batch_write(std::move(mutation_builders), client_state, trace_state, std::move(permit));
     // FIXME: Issue #5650: If we failed writing some of the updates,
     // need to return a list of these failed updates in UnprocessedItems
     // rather than fail the whole write (issue #5650).
@@ -3372,7 +3386,7 @@ static bool hierarchy_filter(rjson::value& val, const attribute_path_map_node<T>
         }
         rjson::value newv = rjson::empty_object();
         for (auto it = v.MemberBegin(); it != v.MemberEnd(); ++it) {
-            std::string attr = it->name.GetString();
+            std::string attr = rjson::to_string(it->name);
             auto x = members.find(attr);
             if (x != members.end()) {
                 if (x->second) {
@@ -3592,7 +3606,7 @@ static std::optional<attrs_to_get> calculate_attrs_to_get(const rjson::value& re
         const rjson::value& attributes_to_get = req["AttributesToGet"];
         attrs_to_get ret;
         for (auto it = attributes_to_get.Begin(); it != attributes_to_get.End(); ++it) {
-            attribute_path_map_add("AttributesToGet", ret, it->GetString());
+            attribute_path_map_add("AttributesToGet", ret, rjson::to_string(*it));
             validate_attr_name_length("AttributesToGet", it->GetStringLength(), false);
         }
         if (ret.empty()) {
@@ -4258,12 +4272,12 @@ inline void update_item_operation::apply_attribute_updates(const std::unique_ptr
         attribute_collector& modified_attrs, bool& any_updates, bool& any_deletes) const {
     for (auto it = _attribute_updates->MemberBegin(); it != _attribute_updates->MemberEnd(); ++it) {
         // Note that it.key() is the name of the column, *it is the operation
-        bytes column_name = to_bytes(it->name.GetString());
+        bytes column_name = to_bytes(rjson::to_string_view(it->name));
         const column_definition* cdef = _schema->get_column_definition(column_name);
         if (cdef && cdef->is_primary_key()) {
-            throw api_error::validation(format("UpdateItem cannot update key column {}", it->name.GetString()));
+            throw api_error::validation(format("UpdateItem cannot update key column {}", rjson::to_string_view(it->name)));
         }
-        std::string action = (it->value)["Action"].GetString();
+        std::string action = rjson::to_string((it->value)["Action"]);
         if (action == "DELETE") {
             // The DELETE operation can do two unrelated tasks. Without a
             // "Value" option, it is used to delete an attribute. With a
@@ -5460,7 +5474,7 @@ calculate_bounds_conditions(schema_ptr schema, const rjson::value& conditions) {
     std::vector<query::clustering_range> ck_bounds;
 
     for (auto it = conditions.MemberBegin(); it != conditions.MemberEnd(); ++it) {
-        std::string key = it->name.GetString();
+        sstring key = rjson::to_sstring(it->name);
         const rjson::value& condition = it->value;
 
         const rjson::value& comp_definition = rjson::get(condition, "ComparisonOperator");
@@ -5468,13 +5482,13 @@ calculate_bounds_conditions(schema_ptr schema, const rjson::value& conditions) {
 
         const column_definition& pk_cdef = schema->partition_key_columns().front();
         const column_definition* ck_cdef = schema->clustering_key_size() > 0 ? &schema->clustering_key_columns().front() : nullptr;
-        if (sstring(key) == pk_cdef.name_as_text()) {
+        if (key == pk_cdef.name_as_text()) {
             if (!partition_ranges.empty()) {
                 throw api_error::validation("Currently only a single restriction per key is allowed");
             }
             partition_ranges.push_back(calculate_pk_bound(schema, pk_cdef, comp_definition, attr_list));
         }
-        if (ck_cdef && sstring(key) == ck_cdef->name_as_text()) {
+        if (ck_cdef && key == ck_cdef->name_as_text()) {
             if (!ck_bounds.empty()) {
                 throw api_error::validation("Currently only a single restriction per key is allowed");
             }
@@ -5875,7 +5889,7 @@ future<executor::request_return_type> executor::list_tables(client_state& client
 
     rjson::value* exclusive_start_json = rjson::find(request, "ExclusiveStartTableName");
     rjson::value* limit_json = rjson::find(request, "Limit");
-    std::string exclusive_start = exclusive_start_json ? exclusive_start_json->GetString() : "";
+    std::string exclusive_start = exclusive_start_json ? rjson::to_string(*exclusive_start_json) : "";
     int limit = limit_json ? limit_json->GetInt() : 100;
     if (limit < 1 || limit > 100) {
         co_return api_error::validation("Limit must be greater than 0 and no greater than 100");

alternator/executor.hh

@@ -40,6 +40,7 @@ namespace cql3::selection {
 
 namespace service {
     class storage_proxy;
+    class cas_shard;
 }
 
 namespace cdc {
@@ -57,6 +58,7 @@ class schema_builder;
 namespace alternator {
 
 class rmw_operation;
+class put_or_delete_item;
 
 schema_ptr get_table(service::storage_proxy& proxy, const rjson::value& request);
 bool is_alternator_keyspace(const sstring& ks_name);
@@ -219,6 +221,16 @@ private:
 
     static void describe_key_schema(rjson::value& parent, const schema&, std::unordered_map<std::string,std::string> * = nullptr, const std::map<sstring, sstring> *tags = nullptr);
 
+    future<> do_batch_write(
+        std::vector<std::pair<schema_ptr, put_or_delete_item>> mutation_builders,
+        service::client_state& client_state,
+        tracing::trace_state_ptr trace_state,
+        service_permit permit);
+
+    future<> cas_write(schema_ptr schema, service::cas_shard cas_shard, const dht::decorated_key& dk,
+        const std::vector<put_or_delete_item>& mutation_builders, service::client_state& client_state,
+        tracing::trace_state_ptr trace_state, service_permit permit);
+
 public:
     static void describe_key_schema(rjson::value& parent, const schema& schema, std::unordered_map<std::string,std::string>&, const std::map<sstring, sstring> *tags = nullptr);
 

alternator/serialization.cc

@@ -496,7 +496,7 @@ const std::pair<std::string, const rjson::value*> unwrap_set(const rjson::value&
         return {"", nullptr};
     }
     auto it = v.MemberBegin();
-    const std::string it_key = it->name.GetString();
+    const std::string it_key = rjson::to_string(it->name);
     if (it_key != "SS" && it_key != "BS" && it_key != "NS") {
         return {std::move(it_key), nullptr};
     }

alternator/ttl.cc

@@ -93,7 +93,7 @@ future<executor::request_return_type> executor::update_time_to_live(client_state
     if (v->GetStringLength() < 1 || v->GetStringLength() > 255) {
         co_return api_error::validation("The length of AttributeName must be between 1 and 255");
     }
-    sstring attribute_name(v->GetString(), v->GetStringLength());
+    sstring attribute_name = rjson::to_sstring(*v);
 
     co_await verify_permission(_enforce_authorization, _warn_authorization, client_state, schema, auth::permission::ALTER, _stats);
     co_await db::modify_tags(_mm, schema->ks_name(), schema->cf_name(), [&](std::map<sstring, sstring>& tags_map) {

api/api-doc/storage_service.json

@@ -3051,7 +3051,7 @@
                   },
                   {
                      "name":"incremental_mode",
-                     "description":"Set the incremental repair mode. Can be 'disabled', 'incremental', or 'full'. 'incremental': The incremental repair logic is enabled. Unrepaired sstables will be included for repair. Repaired sstables will be skipped. The incremental repair states will be updated after repair. 'full': The incremental repair logic is enabled. Both repaired and unrepaired sstables will be included for repair. The incremental repair states will be updated after repair. 'disabled': The incremental repair logic is disabled completely. The incremental repair states, e.g., repaired_at in sstables and sstables_repaired_at in the system.tablets table, will not be updated after repair. When the option is not provided, it defaults to incremental mode.",
+                     "description":"Set the incremental repair mode. Can be 'disabled', 'incremental', or 'full'. 'incremental': The incremental repair logic is enabled. Unrepaired sstables will be included for repair. Repaired sstables will be skipped. The incremental repair states will be updated after repair. 'full': The incremental repair logic is enabled. Both repaired and unrepaired sstables will be included for repair. The incremental repair states will be updated after repair. 'disabled': The incremental repair logic is disabled completely. The incremental repair states, e.g., repaired_at in sstables and sstables_repaired_at in the system.tablets table, will not be updated after repair. When the option is not provided, it defaults to 'disabled' mode.",
                      "required":false,
                      "allowMultiple":false,
                      "type":"string",

auth/certificate_authenticator.cc

@@ -77,9 +77,9 @@ auth::certificate_authenticator::certificate_authenticator(cql3::query_processor
                         throw std::invalid_argument(fmt::format("Invalid source: {}", map.at(cfg_source_attr)));
                     }
                     continue;
-                } catch (std::out_of_range&) {
+                } catch (const std::out_of_range&) {
                     // just fallthrough
-                } catch (boost::regex_error&) {
+                } catch (const boost::regex_error&) {
                     std::throw_with_nested(std::invalid_argument(fmt::format("Invalid query expression: {}", map.at(cfg_query_attr))));
                 }
             }

auth/common.cc

@@ -94,7 +94,7 @@ static future<> create_legacy_metadata_table_if_missing_impl(
         try {
             co_return co_await mm.announce(co_await ::service::prepare_new_column_family_announcement(qp.proxy(), table, ts),
                     std::move(group0_guard), format("auth: create {} metadata table", table->cf_name()));
-        } catch (exceptions::already_exists_exception&) {}
+        } catch (const exceptions::already_exists_exception&) {}
     }
 }
 

auth/default_authorizer.cc

@@ -256,7 +256,7 @@ future<> default_authorizer::revoke_all(std::string_view role_name, ::service::g
         } else {
             co_await collect_mutations(_qp, mc, query, {sstring(role_name)});
         }
-    } catch (exceptions::request_execution_exception& e) {
+    } catch (const exceptions::request_execution_exception& e) {
         alogger.warn("CassandraAuthorizer failed to revoke all permissions of {}: {}", role_name, e);
     }
 }
@@ -293,13 +293,13 @@ future<> default_authorizer::revoke_all_legacy(const resource& resource) {
                                 [resource](auto ep) {
                     try {
                         std::rethrow_exception(ep);
-                    } catch (exceptions::request_execution_exception& e) {
+                    } catch (const exceptions::request_execution_exception& e) {
                         alogger.warn("CassandraAuthorizer failed to revoke all permissions on {}: {}", resource, e);
                     }
 
                 });
             });
-        } catch (exceptions::request_execution_exception& e) {
+        } catch (const exceptions::request_execution_exception& e) {
             alogger.warn("CassandraAuthorizer failed to revoke all permissions on {}: {}", resource, e);
             return make_ready_future();
         }

auth/password_authenticator.cc

@@ -337,13 +337,13 @@ future<authenticated_user> password_authenticator::authenticate(
             throw exceptions::authentication_exception("Username and/or password are incorrect");
         }
         co_return username;
-    } catch (std::system_error &) {
+    } catch (const std::system_error &) {
         std::throw_with_nested(exceptions::authentication_exception("Could not verify password"));
-    } catch (exceptions::request_execution_exception& e) {
+    } catch (const exceptions::request_execution_exception& e) {
         std::throw_with_nested(exceptions::authentication_exception(e.what()));
-    } catch (exceptions::authentication_exception& e) {
+    } catch (const exceptions::authentication_exception& e) {
         std::throw_with_nested(e);
-    } catch (exceptions::unavailable_exception& e) {
+    } catch (const exceptions::unavailable_exception& e) {
         std::throw_with_nested(exceptions::authentication_exception(e.get_message()));
     } catch (...) {
         std::throw_with_nested(exceptions::authentication_exception("authentication failed"));

auth/service.cc

@@ -226,7 +226,7 @@ future<> service::create_legacy_keyspace_if_missing(::service::migration_manager
             try {
                 co_return co_await mm.announce(::service::prepare_new_keyspace_announcement(db.real_database(), ksm, ts),
                         std::move(group0_guard), seastar::format("auth_service: create {} keyspace", meta::legacy::AUTH_KS));
-            } catch (::service::group0_concurrent_modification&) {
+            } catch (const ::service::group0_concurrent_modification&) {
                 log.info("Concurrent operation is detected while creating {} keyspace, retrying.", meta::legacy::AUTH_KS);
             }
         }

auth/standard_role_manager.cc

@@ -192,7 +192,7 @@ future<> standard_role_manager::legacy_create_default_role_if_missing() {
                 {_superuser},
                 cql3::query_processor::cache_internal::no).discard_result();
         log.info("Created default superuser role '{}'.", _superuser);
-    } catch(const exceptions::unavailable_exception& e) {
+    } catch (const exceptions::unavailable_exception& e) {
         log.warn("Skipped default role setup: some nodes were not ready; will retry");
         throw e;
     }

auth/transitional.cc

@@ -81,7 +81,7 @@ public:
         }).handle_exception([](auto ep) {
             try {
                 std::rethrow_exception(ep);
-            } catch (exceptions::authentication_exception&) {
+            } catch (const exceptions::authentication_exception&) {
                 // return anon user
                 return make_ready_future<authenticated_user>(anonymous_user());
             }
@@ -126,7 +126,7 @@ public:
             virtual bytes evaluate_response(bytes_view client_response) override {
                 try {
                     return _sasl->evaluate_response(client_response);
-                } catch (exceptions::authentication_exception&) {
+                } catch (const exceptions::authentication_exception&) {
                     _complete = true;
                     return {};
                 }
@@ -141,7 +141,7 @@ public:
                     return _sasl->get_authenticated_user().handle_exception([](auto ep) {
                         try {
                             std::rethrow_exception(ep);
-                        } catch (exceptions::authentication_exception&) {
+                        } catch (const exceptions::authentication_exception&) {
                             // return anon user
                             return make_ready_future<authenticated_user>(anonymous_user());
                         }

db/view/view.cc

@@ -1744,6 +1744,115 @@ bool should_generate_view_updates_on_this_shard(const schema_ptr& base, const lo
         && std::ranges::contains(shards, this_shard_id());
 }
 
+static endpoints_to_update get_view_natural_endpoint_vnodes(
+        locator::host_id me,
+        std::vector<std::reference_wrapper<const locator::node>> base_nodes,
+        std::vector<std::reference_wrapper<const locator::node>> view_nodes,
+        locator::endpoint_dc_rack my_location,
+        const locator::network_topology_strategy* network_topology,
+        replica::cf_stats& cf_stats) {
+    using node_vector = std::vector<std::reference_wrapper<const locator::node>>;
+    node_vector base_endpoints, view_endpoints;
+    auto& my_datacenter = my_location.dc;
+
+    auto process_candidate = [&] (node_vector& nodes, std::reference_wrapper<const locator::node> node) {
+        if (!network_topology || node.get().dc() == my_datacenter) {
+            nodes.emplace_back(node);
+        }
+    };
+
+    for (auto&& base_node : base_nodes) {
+        process_candidate(base_endpoints, base_node);
+    }
+
+    for (auto&& view_node : view_nodes) {
+        auto it = std::ranges::find(base_endpoints, view_node.get().host_id(), std::mem_fn(&locator::node::host_id));
+        // If this base replica is also one of the view replicas, we use
+        // ourselves as the view replica.
+        // We don't return an extra endpoint, as it's only needed when
+        // using tablets (so !use_legacy_self_pairing)
+        if (view_node.get().host_id() == me && it != base_endpoints.end()) {
+            return {.natural_endpoint = me};
+        }
+
+        // We have to remove any endpoint which is shared between the base
+        // and the view, as it will select itself and throw off the counts
+        // otherwise.
+        if (it != base_endpoints.end()) {
+            base_endpoints.erase(it);
+        } else if (!network_topology || view_node.get().dc() == my_datacenter) {
+            view_endpoints.push_back(view_node);
+        }
+    }
+
+    auto base_it = std::ranges::find(base_endpoints, me, std::mem_fn(&locator::node::host_id));
+    if (base_it == base_endpoints.end()) {
+        // This node is not a base replica of this key, so we return empty
+        // FIXME: This case shouldn't happen, and if it happens, a view update
+        // would be lost.
+        ++cf_stats.total_view_updates_on_wrong_node;
+        vlogger.warn("Could not find {} in base_endpoints={}", me,
+                base_endpoints | std::views::transform(std::mem_fn(&locator::node::host_id)));
+        return {};
+    }
+    size_t idx = base_it - base_endpoints.begin();
+    return {.natural_endpoint = view_endpoints[idx].get().host_id()};
+}
+
+static std::optional<locator::host_id> get_unpaired_view_endpoint(
+        std::vector<std::reference_wrapper<const locator::node>> base_nodes,
+        std::vector<std::reference_wrapper<const locator::node>> view_nodes,
+        replica::cf_stats& cf_stats) {
+    std::unordered_set<locator::endpoint_dc_rack> base_dc_racks;
+    for (auto&& base_node : base_nodes) {
+        if (base_dc_racks.contains(base_node.get().dc_rack())) {
+            // We can't do rack-aware pairing if there are multiple replicas in the same rack.
+            ++cf_stats.total_view_updates_failed_pairing;
+            vlogger.warn("Can't perform base-view pairing in this topology. There are multiple base table replicas in the same dc/rack({}/{}):",
+                    base_node.get().dc(), base_node.get().rack());
+            return std::nullopt;
+        }
+        base_dc_racks.insert(base_node.get().dc_rack());
+    }
+
+    std::unordered_set<locator::endpoint_dc_rack> paired_view_dc_racks;
+    std::unordered_map<locator::endpoint_dc_rack, locator::host_id> unpaired_view_dc_rack_replicas;
+    for (auto&& view_node : view_nodes) {
+        if (paired_view_dc_racks.contains(view_node.get().dc_rack()) || unpaired_view_dc_rack_replicas.contains(view_node.get().dc_rack())) {
+            // We can't do rack-aware pairing if there are multiple replicas in the same rack.
+            ++cf_stats.total_view_updates_failed_pairing;
+            vlogger.warn("Can't perform base-view pairing in this topology. There are multiple view table replicas in the same dc/rack({}/{}):",
+                    view_node.get().dc(), view_node.get().rack());
+            return std::nullopt;
+        }
+        // Track unpaired replicas in both sets
+        if (base_dc_racks.contains(view_node.get().dc_rack())) {
+            paired_view_dc_racks.insert(view_node.get().dc_rack());
+        } else {
+            unpaired_view_dc_rack_replicas.insert({view_node.get().dc_rack(), view_node.get().host_id()});
+        }
+    }
+
+    if (unpaired_view_dc_rack_replicas.size() > 0) {
+        // There are view replicas that can't be paired with any base replica
+        // This can happen as a result of an RF change when the view replica finishes streaming
+        // before the base replica.
+        // Because of this, a view replica might not get paired with any base replica, so we need
+        // to send an additional update to it.
+        ++cf_stats.total_view_updates_due_to_replica_count_mismatch;
+        auto extra_replica = unpaired_view_dc_rack_replicas.begin()->second;
+        unpaired_view_dc_rack_replicas.erase(unpaired_view_dc_rack_replicas.begin());
+        if (unpaired_view_dc_rack_replicas.size() > 0) {
+            // We only expect one extra replica to appear due to an RF change. If there's more, that's an error,
+            // but we'll still perform updates to the paired and last replicas to minimize degradation.
+            vlogger.warn("There are too many view endpoints for base-view pairing. View updates may get lost on view_endpoints={}",
+                    unpaired_view_dc_rack_replicas | std::views::values);
+        }
+        return extra_replica;
+    }
+    return std::nullopt;
+}
+
 // Calculate the node ("natural endpoint") to which this node should send
 // a view update.
 //
@@ -1756,29 +1865,19 @@ bool should_generate_view_updates_on_this_shard(const schema_ptr& base, const lo
 // of this function is to find, assuming that this node is one of the base
 // replicas for a given partition, the paired view replica.
 //
-// In the past, we used an optimization called "self-pairing" that if a single
-// node was both a base replica and a view replica for a write, the pairing is
-// modified so that this node would send the update to itself. This self-
-// pairing optimization could cause the pairing to change after view ranges
-// are moved between nodes, so currently we only use it if
-// use_legacy_self_pairing is set to true. When using tablets - where range
-// movements are common - it is strongly recommended to set it to false.
+// When using vnodes, we have an optimization called "self-pairing" - if a single
+// node is both a base replica and a view replica for a write, the pairing is
+// modified so that this node sends the update to itself and this node is removed
+// from the lists of nodes paired by index. This self-pairing optimization can
+// cause the pairing to change after view ranges are moved between nodes.
 //
 // If the keyspace's replication strategy is a NetworkTopologyStrategy,
 // we pair only nodes in the same datacenter.
 //
-// When use_legacy_self_pairing is enabled, if one of the base replicas
-// also happens to be a view replica, it is paired with itself
-// (with the other nodes paired by order in the list
-// after taking this node out).
-//
-// If the table uses tablets and the replication strategy is NetworkTopologyStrategy
-// and the replication factor in the node's datacenter is a multiple of the number
-// of racks in the datacenter, then pairing is rack-aware.  In this case,
-// all racks have the same number of replicas, and those are never migrated
-// outside their racks. Therefore, the base replicas are naturally paired with the
-// view replicas that are in the same rack, based on the ordinal position.
-// Note that typically, there is a single replica per rack and pairing is trivial.
+// If the table uses tablets, then pairing is rack-aware. In this case, in each
+// rack where we have a base replica there is also one replica of each view tablet.
+// Therefore, the base replicas are naturally paired with the view replicas that
+// are in the same rack.
 //
 // If the assumption that the given base token belongs to this replica
 // does not hold, we return an empty optional.
@@ -1806,19 +1905,12 @@ endpoints_to_update get_view_natural_endpoint(
         const locator::abstract_replication_strategy& replication_strategy,
         const dht::token& base_token,
         const dht::token& view_token,
-        bool use_legacy_self_pairing,
-        bool use_tablets_rack_aware_view_pairing,
+        bool use_tablets,
         replica::cf_stats& cf_stats) {
     auto& topology = base_erm->get_token_metadata_ptr()->get_topology();
     auto& view_topology = view_erm->get_token_metadata_ptr()->get_topology();
     auto& my_location = topology.get_location(me);
-    auto& my_datacenter = my_location.dc;
     auto* network_topology = dynamic_cast<const locator::network_topology_strategy*>(&replication_strategy);
-    auto rack_aware_pairing = use_tablets_rack_aware_view_pairing && network_topology;
-    bool simple_rack_aware_pairing = false;
-    using node_vector = std::vector<std::reference_wrapper<const locator::node>>;
-    node_vector orig_base_endpoints, orig_view_endpoints;
-    node_vector base_endpoints, view_endpoints;
 
     auto resolve = [&] (const locator::topology& topology, const locator::host_id& ep, bool is_view) -> const locator::node& {
         if (auto* np = topology.find_node(ep)) {
@@ -1829,6 +1921,7 @@ endpoints_to_update get_view_natural_endpoint(
 
     // We need to use get_replicas() for pairing to be stable in case base or view tablet
     // is rebuilding a replica which has left the ring. get_natural_endpoints() filters such replicas.
+    using node_vector = std::vector<std::reference_wrapper<const locator::node>>;
     auto base_nodes = base_erm->get_replicas(base_token) | std::views::transform([&] (const locator::host_id& ep) -> const locator::node& {
         return resolve(topology, ep, false);
     }) | std::ranges::to<node_vector>();
@@ -1852,231 +1945,43 @@ endpoints_to_update get_view_natural_endpoint(
                 // note that the recursive call will not recurse again because leaving_base is in base_nodes.
                 auto leaving_base = it->get().host_id();
                 return get_view_natural_endpoint(leaving_base, base_erm, view_erm, replication_strategy, base_token,
-                        view_token, use_legacy_self_pairing, use_tablets_rack_aware_view_pairing, cf_stats);
+                        view_token, use_tablets, cf_stats);
             }
         }
     }
 
-    std::function<bool(const locator::node&)> is_candidate;
-    if (network_topology) {
-        is_candidate = [&] (const locator::node& node) { return node.dc() == my_datacenter; };
-    } else {
-        is_candidate = [&] (const locator::node&) { return true; };
-    }
-    auto process_candidate = [&] (node_vector& nodes, std::reference_wrapper<const locator::node> node) {
-        if (is_candidate(node)) {
-            nodes.emplace_back(node);
-        }
-    };
-
-    for (auto&& base_node : base_nodes) {
-        process_candidate(base_endpoints, base_node);
+    if (!use_tablets) {
+        return get_view_natural_endpoint_vnodes(
+                me,
+                base_nodes,
+                view_nodes,
+                my_location,
+                network_topology,
+                cf_stats);
     }
 
-    if (use_legacy_self_pairing) {
-        for (auto&& view_node : view_nodes) {
-            auto it = std::ranges::find(base_endpoints, view_node.get().host_id(), std::mem_fn(&locator::node::host_id));
-            // If this base replica is also one of the view replicas, we use
-            // ourselves as the view replica.
-            // We don't return an extra endpoint, as it's only needed when
-            // using tablets (so !use_legacy_self_pairing)
-            if (view_node.get().host_id() == me && it != base_endpoints.end()) {
-                return {.natural_endpoint = me};
-            }
-
-            // We have to remove any endpoint which is shared between the base
-            // and the view, as it will select itself and throw off the counts
-            // otherwise.
-            if (it != base_endpoints.end()) {
-                base_endpoints.erase(it);
-            } else if (is_candidate(view_node)) {
-                view_endpoints.push_back(view_node);
-            }
-        }
-    } else {
-        for (auto&& view_node : view_nodes) {
-            process_candidate(view_endpoints, view_node);
+    std::optional<locator::host_id> paired_replica;
+    for (auto&& view_node : view_nodes) {
+        if (view_node.get().dc_rack() == my_location) {
+            paired_replica = view_node.get().host_id();
+            break;
         }
     }
-
-    // Try optimizing for simple rack-aware pairing
-    // If the numbers of base and view replica differ, that means an RF change is taking place
-    // and we can't use simple rack-aware pairing.
-    if (rack_aware_pairing && base_endpoints.size() == view_endpoints.size()) {
-        auto dc_rf = network_topology->get_replication_factor(my_datacenter);
-        const auto& racks = topology.get_datacenter_rack_nodes().at(my_datacenter);
-        // Simple rack-aware pairing is possible when the datacenter replication factor
-        // is a multiple of the number of racks in the datacenter.
-        if (dc_rf % racks.size() == 0) {
-            simple_rack_aware_pairing = true;
-            size_t rack_rf = dc_rf / racks.size();
-            // If any rack doesn't have enough nodes to satisfy the per-rack rf
-            // simple rack-aware pairing is disabled.
-            for (const auto& [rack, nodes] : racks) {
-                if (nodes.size() < rack_rf) {
-                    simple_rack_aware_pairing = false;
-                    break;
-                }
-            }
-        }
-        if (dc_rf != base_endpoints.size()) {
-            // If the datacenter replication factor is not equal to the number of base replicas,
-            // we're in progress of a RF change and we can't use simple rack-aware pairing.
-            simple_rack_aware_pairing = false;
-        }
-        if (simple_rack_aware_pairing) {
-            std::erase_if(base_endpoints, [&] (const locator::node& node) { return node.dc_rack() != my_location; });
-            std::erase_if(view_endpoints, [&] (const locator::node& node) { return node.dc_rack() != my_location; });
-        }
+    if (paired_replica && base_nodes.size() == view_nodes.size()) {
+        // We don't need to find any extra replicas, so we can return early
+        return {.natural_endpoint = paired_replica};
     }
-
-    orig_base_endpoints = base_endpoints;
-    orig_view_endpoints = view_endpoints;
-
-    // For the complex rack_aware_pairing case, nodes are already filtered by datacenter
-    // Use best-match, for the minimum number of base and view replicas in each rack,
-    // and ordinal match for the rest.
-    std::optional<std::reference_wrapper<const locator::node>> paired_replica;
-    if (rack_aware_pairing && !simple_rack_aware_pairing) {
-        struct indexed_replica {
-            size_t idx;
-            std::reference_wrapper<const locator::node> node;
-        };
-        std::unordered_map<sstring, std::vector<indexed_replica>> base_racks, view_racks;
-
-        // First, index all replicas by rack
-        auto index_replica_set = [] (std::unordered_map<sstring, std::vector<indexed_replica>>& racks, const node_vector& replicas) {
-            size_t idx = 0;
-            for (const auto& r: replicas) {
-                racks[r.get().rack()].emplace_back(idx++, r);
-            }
-        };
-        index_replica_set(base_racks, base_endpoints);
-        index_replica_set(view_racks, view_endpoints);
-
-        // Try optimistically pairing `me` first
-        const auto& my_base_replicas = base_racks[my_location.rack];
-        auto base_it = std::ranges::find(my_base_replicas, me, [] (const indexed_replica& ir) { return ir.node.get().host_id(); });
-        if (base_it == my_base_replicas.end()) {
-            return {};
-        }
-        const auto& my_view_replicas = view_racks[my_location.rack];
-        size_t idx = base_it - my_base_replicas.begin();
-        if (idx < my_view_replicas.size()) {
-            if (orig_view_endpoints.size() <= orig_base_endpoints.size()) {
-                return {.natural_endpoint = my_view_replicas[idx].node.get().host_id()};
-            } else {
-                // If the number of view replicas is larger than the number of base replicas,
-                // we need to find the unpaired view replica, so we can't return yet.
-                paired_replica = my_view_replicas[idx].node;
-            }
-        }
-
-        // Collect all unpaired base and view replicas,
-        // where the number of replicas in the base rack is different than the respective view rack
-        std::vector<indexed_replica> unpaired_base_replicas, unpaired_view_replicas;
-        for (const auto& [rack, base_replicas] : base_racks) {
-            const auto& view_replicas = view_racks[rack];
-            for (auto i = view_replicas.size(); i < base_replicas.size(); ++i) {
-                unpaired_base_replicas.emplace_back(base_replicas[i]);
-            }
-        }
-        for (const auto& [rack, view_replicas] : view_racks) {
-            const auto& base_replicas = base_racks[rack];
-            for (auto i = base_replicas.size(); i < view_replicas.size(); ++i) {
-                unpaired_view_replicas.emplace_back(view_replicas[i]);
-            }
-        }
-
-        // Sort by the original ordinality, and copy the sorted results
-        // back into {base,view}_endpoints, for backward compatible processing below.
-        std::ranges::sort(unpaired_base_replicas, std::less(), std::mem_fn(&indexed_replica::idx));
-        base_endpoints.clear();
-        std::ranges::transform(unpaired_base_replicas, std::back_inserter(base_endpoints), std::mem_fn(&indexed_replica::node));
-
-        std::ranges::sort(unpaired_view_replicas, std::less(), std::mem_fn(&indexed_replica::idx));
-        view_endpoints.clear();
-        std::ranges::transform(unpaired_view_replicas, std::back_inserter(view_endpoints), std::mem_fn(&indexed_replica::node));
-    }
-
-    auto base_it = std::ranges::find(base_endpoints, me, std::mem_fn(&locator::node::host_id));
-    if (!paired_replica && base_it == base_endpoints.end()) {
-        // This node is not a base replica of this key, so we return empty
-        // FIXME: This case shouldn't happen, and if it happens, a view update
-        // would be lost.
-        ++cf_stats.total_view_updates_on_wrong_node;
-        vlogger.warn("Could not find {} in base_endpoints={}", me,
-                orig_base_endpoints | std::views::transform(std::mem_fn(&locator::node::host_id)));
-        return {};
-    }
-    size_t idx = base_it - base_endpoints.begin();
-    std::optional<std::reference_wrapper<const locator::node>> no_pairing_replica;
-    if (!paired_replica && idx >= view_endpoints.size()) {
-        // There are fewer view replicas than base replicas
-        // FIXME: This might still happen when reducing replication factor with tablets,
-        // see https://github.com/scylladb/scylladb/issues/21492
-        ++cf_stats.total_view_updates_failed_pairing;
-        vlogger.warn("Could not pair {}: rack_aware={} base_endpoints={} view_endpoints={}", me,
-                rack_aware_pairing ? (simple_rack_aware_pairing ? "simple" : "complex") : "none",
-                orig_base_endpoints | std::views::transform(std::mem_fn(&locator::node::host_id)),
-                orig_view_endpoints | std::views::transform(std::mem_fn(&locator::node::host_id)));
-        return {};
-    } else if (base_endpoints.size() < view_endpoints.size()) {
-        // There are fewer base replicas than view replicas.
-        // This can happen as a result of an RF change when the view replica finishes streaming
-        // before the base replica.
-        // Because of this, a view replica might not get paired with any base replica, so we need
-        // to send an additional update to it.
-        ++cf_stats.total_view_updates_due_to_replica_count_mismatch;
-        no_pairing_replica = view_endpoints.back();
-        if (base_endpoints.size() < view_endpoints.size() - 1) {
-            // We only expect one extra replica to appear due to an RF change. If there's more, that's an error,
-            // but we'll still perform updates to the paired and last replicas to minimize degradation.
-            vlogger.warn("There are too many view endpoints for base-view pairing. View updates may get lost on view_endpoints={}",
-                    std::span(view_endpoints.begin() + base_endpoints.size(), view_endpoints.end() - 1) | std::views::transform(std::mem_fn(&locator::node::host_id)));
-        }
-    }
-
     if (!paired_replica) {
-        paired_replica = view_endpoints[idx];
+        // We couldn't find any view replica in our rack
+        ++cf_stats.total_view_updates_failed_pairing;
+        vlogger.warn("Could not find a view replica in the same rack as base replica {} for base_endpoints={} view_endpoints={}", 
+                me,
+                base_nodes | std::views::transform(std::mem_fn(&locator::node::host_id)),
+                view_nodes | std::views::transform(std::mem_fn(&locator::node::host_id)));
     }
-    if (!no_pairing_replica && base_nodes.size() < view_nodes.size()) {
-        // This can happen when the view replica with no pairing is in another DC.
-        // We need to send an update to it if there are no base replicas in that DC yet,
-        // as it won't receive updates otherwise.
-        std::unordered_set<sstring> dcs_with_base_replicas;
-        for (const auto& base_node : base_nodes) {
-            dcs_with_base_replicas.insert(base_node.get().dc());
-        }
-        for (const auto& view_node : view_nodes) {
-            if (!dcs_with_base_replicas.contains(view_node.get().dc())) {
-                ++cf_stats.total_view_updates_due_to_replica_count_mismatch;
-                no_pairing_replica = view_node;
-                break;
-            }
-        }
-    }
-    // https://github.com/scylladb/scylladb/issues/19439
-    // With tablets, a node being replaced might transition to "left" state
-    // but still be kept as a replica.
-    // As of writing this hints are not prepared to handle nodes that are left
-    // but are still replicas. Therefore, there is no other sensible option
-    // right now but to give up attempt to send the update or write a hint
-    // to the paired, permanently down replica.
-    // We use the same workaround for the extra replica.
-    auto return_host_id_if_not_left = [] (const auto& replica) -> std::optional<locator::host_id> {
-        if (!replica) {
-            return std::nullopt;
-        }
-        const auto& node = replica->get();
-        if (!node.left()) {
-            return node.host_id();
-        } else {
-            return std::nullopt;
-        }
-    };
-    return {.natural_endpoint = return_host_id_if_not_left(paired_replica),
-            .endpoint_with_no_pairing = return_host_id_if_not_left(no_pairing_replica)};
+    std::optional<locator::host_id> no_pairing_replica = get_unpaired_view_endpoint(base_nodes, view_nodes, cf_stats);
+    return {.natural_endpoint = paired_replica,
+            .endpoint_with_no_pairing = no_pairing_replica};
 }
 
 static future<> apply_to_remote_endpoints(service::storage_proxy& proxy, locator::effective_replication_map_ptr ermp,
@@ -2136,12 +2041,6 @@ future<> view_update_generator::mutate_MV(
 {
     auto& ks = _db.find_keyspace(base->ks_name());
     auto& replication = ks.get_replication_strategy();
-    // We set legacy self-pairing for old vnode-based tables (for backward
-    // compatibility), and unset it for tablets - where range movements
-    // are more frequent and backward compatibility is less important.
-    // TODO: Maybe allow users to set use_legacy_self_pairing explicitly
-    // on a view, like we have the synchronous_updates_flag.
-    bool use_legacy_self_pairing = !ks.uses_tablets();
     std::unordered_map<table_id, locator::effective_replication_map_ptr> erms;
     auto get_erm = [&] (table_id id) {
         auto it = erms.find(id);
@@ -2154,10 +2053,6 @@ future<> view_update_generator::mutate_MV(
     for (const auto& mut : view_updates) {
         (void)get_erm(mut.s->id());
     }
-    // Enable rack-aware view updates pairing for tablets
-    // when the cluster feature is enabled so that all replicas agree
-    // on the pairing algorithm.
-    bool use_tablets_rack_aware_view_pairing = _db.features().tablet_rack_aware_view_pairing && ks.uses_tablets();
     auto me = base_ermp->get_topology().my_host_id();
     static constexpr size_t max_concurrent_updates = 128;
     co_await utils::get_local_injector().inject("delay_before_get_view_natural_endpoint", 8000ms);
@@ -2165,7 +2060,7 @@ future<> view_update_generator::mutate_MV(
         auto view_token = dht::get_token(*mut.s, mut.fm.key());
         auto view_ermp = erms.at(mut.s->id());
         auto [target_endpoint, no_pairing_endpoint] = get_view_natural_endpoint(me, base_ermp, view_ermp, replication, base_token, view_token,
-                use_legacy_self_pairing, use_tablets_rack_aware_view_pairing, cf_stats);
+                ks.uses_tablets(), cf_stats);
         auto remote_endpoints = view_ermp->get_pending_replicas(view_token);
         auto memory_units = seastar::make_lw_shared<db::timeout_semaphore_units>(pending_view_update_memory_units.split(memory_usage_of(mut)));
         if (no_pairing_endpoint) {

db/view/view.hh

@@ -305,8 +305,7 @@ endpoints_to_update get_view_natural_endpoint(
     const locator::abstract_replication_strategy& replication_strategy,
     const dht::token& base_token,
     const dht::token& view_token,
-    bool use_legacy_self_pairing,
-    bool use_tablets_basic_rack_aware_view_pairing,
+    bool use_tablets,
     replica::cf_stats& cf_stats);
 
 /// Verify that the provided keyspace is eligible for storing materialized views.

dist/common/sysconfig/scylla-node-exporter

@@ -1 +1 @@
-SCYLLA_NODE_EXPORTER_ARGS="--collector.interrupts --no-collector.hwmon --no-collector.bcache --no-collector.btrfs --no-collector.fibrechannel --no-collector.infiniband --no-collector.ipvs --no-collector.nfs --no-collector.nfsd --no-collector.powersupplyclass --no-collector.rapl --no-collector.tapestats --no-collector.thermal_zone --no-collector.udp_queues --no-collector.zfs"
+SCYLLA_NODE_EXPORTER_ARGS="--collector.interrupts --collector.ethtool.metrics-include='(bw_in_allowance_exceeded|bw_out_allowance_exceeded|conntrack_allowance_exceeded|conntrack_allowance_available|linklocal_allowance_exceeded)' --collector.ethtool --no-collector.hwmon --no-collector.bcache --no-collector.btrfs --no-collector.fibrechannel --no-collector.infiniband --no-collector.ipvs --no-collector.nfs --no-collector.nfsd --no-collector.powersupplyclass --no-collector.rapl --no-collector.tapestats --no-collector.thermal_zone --no-collector.udp_queues --no-collector.zfs"

docs/features/incremental-repair.rst

@@ -28,7 +28,8 @@ Incremental Repair is only supported for tables that use the tablets architectur
 Incremental Repair Modes
 ------------------------
 
-While incremental repair is the default and recommended mode, you can control its behavior for a given repair operation using the ``incremental_mode`` parameter. This is useful for situations where you might need to force a full data validation.
+Incremental is currently disabled by default. You can control its behavior for a given repair operation using the ``incremental_mode`` parameter.
+This is useful for enabling incremental repair, or in situations where you might need to force a full data validation.
 
 The available modes are:
 

docs/operating-scylla/nodetool-commands/cluster/repair.rst

@@ -53,13 +53,13 @@ ScyllaDB nodetool cluster repair command supports the following options:
 
      nodetool cluster repair --tablet-tokens 1,10474535988
 
-- ``--incremental-mode`` specifies the incremental repair mode. Can be 'disabled', 'incremental', or 'full'. 'incremental': The incremental repair logic is enabled. Unrepaired sstables will be included for repair. Repaired sstables will be skipped. The incremental repair states will be updated after repair. 'full': The incremental repair logic is enabled. Both repaired and unrepaired sstables will be included for repair. The incremental repair states will be updated after repair. 'disabled': The incremental repair logic is disabled completely. The incremental repair states, e.g., repaired_at in sstables and sstables_repaired_at in the system.tablets table, will not be updated after repair. When the option is not provided, it defaults to incremental.
+- ``--incremental-mode`` specifies the incremental repair mode. Can be 'disabled', 'incremental', or 'full'. 'incremental': The incremental repair logic is enabled. Unrepaired sstables will be included for repair. Repaired sstables will be skipped. The incremental repair states will be updated after repair. 'full': The incremental repair logic is enabled. Both repaired and unrepaired sstables will be included for repair. The incremental repair states will be updated after repair. 'disabled': The incremental repair logic is disabled completely. The incremental repair states, e.g., repaired_at in sstables and sstables_repaired_at in the system.tablets table, will not be updated after repair. When the option is not provided, it defaults to 'disabled'.
 
   For example:
 
   ::
 
-     nodetool cluster repair --incremental-mode regular
+     nodetool cluster repair --incremental-mode disabled
 
 - ``keyspace`` executes a repair on a specific keyspace. The default is all keyspaces.
 

docs/poetry.lock

@@ -1018,14 +1018,14 @@ sphinx-markdown-tables = "0.0.17"
 
 [[package]]
 name = "sphinx-scylladb-theme"
-version = "1.8.9"
+version = "1.8.10"
 description = "A Sphinx Theme for ScyllaDB documentation projects"
 optional = false
 python-versions = "<4.0,>=3.10"
 groups = ["main"]
 files = [
-    {file = "sphinx_scylladb_theme-1.8.9-py3-none-any.whl", hash = "sha256:f8649a7753a29494fd2b417d1cb855035dddb9ebd498ea033fd73f5f9338271e"},
-    {file = "sphinx_scylladb_theme-1.8.9.tar.gz", hash = "sha256:ab7cda4c10a0d067c5c3a45f7b1f68cb8ebefe135a0be0738bfa282a344769b6"},
+    {file = "sphinx_scylladb_theme-1.8.10-py3-none-any.whl", hash = "sha256:8b930f33bec7308ccaa92698ebb5ad85059bcbf93a463f92917aeaf473fce632"},
+    {file = "sphinx_scylladb_theme-1.8.10.tar.gz", hash = "sha256:8a78a9b692d9a946be2c4a64aa472fd82204cc8ea0b1ee7f60de6db35b356326"},
 ]
 
 [package.dependencies]
@@ -1603,4 +1603,4 @@ files = [
 [metadata]
 lock-version = "2.1"
 python-versions = "^3.10"
-content-hash = "74912627a3f424290ed7889451c0bdb1a862ab85b1d07c85f4f3b8c34f32a020"
+content-hash = "0ae673106f45d3465cbdabbf511e165ca44feadd34d7753f2e68093afaa95c79"

docs/pyproject.toml

@@ -9,7 +9,7 @@ package-mode = false
 python = "^3.10"
 pygments = "^2.18.0"
 redirects_cli ="^0.1.3"
-sphinx-scylladb-theme = "^1.8.9"
+sphinx-scylladb-theme = "^1.8.10"
 sphinx-sitemap = "^2.6.0"
 sphinx-autobuild = "^2024.4.19"
 Sphinx = "^7.3.7"

install-dependencies.sh

@@ -38,6 +38,7 @@ debian_base_packages=(
     python3-aiohttp
     python3-pyparsing
     python3-colorama
+    python3-dev
     python3-tabulate
     python3-pytest
     python3-pytest-asyncio
@@ -65,6 +66,7 @@ debian_base_packages=(
     git-lfs
     e2fsprogs
     fuse3
+    libev-dev # for python driver
 )
 
 fedora_packages=(
@@ -90,6 +92,7 @@ fedora_packages=(
     patchelf
     python3
     python3-aiohttp
+    python3-devel
     python3-pip
     python3-file-magic
     python3-colorama
@@ -154,6 +157,8 @@ fedora_packages=(
     https://github.com/scylladb/cassandra-stress/releases/download/v3.18.1/cassandra-stress-java21-3.18.1-1.noarch.rpm
     elfutils
     jq
+
+    libev-devel # for python driver
 )
 
 fedora_python3_packages=(

locator/tablets.hh

@@ -200,7 +200,10 @@ enum class tablet_repair_incremental_mode : uint8_t {
     disabled,
 };
 
-constexpr tablet_repair_incremental_mode default_tablet_repair_incremental_mode{tablet_repair_incremental_mode::incremental};
+// FIXME: Incremental repair is disabled by default due to
+// https://github.com/scylladb/scylladb/issues/26041 and
+// https://github.com/scylladb/scylladb/issues/27414
+constexpr tablet_repair_incremental_mode default_tablet_repair_incremental_mode{tablet_repair_incremental_mode::disabled};
 
 sstring tablet_repair_incremental_mode_to_string(tablet_repair_incremental_mode);
 tablet_repair_incremental_mode tablet_repair_incremental_mode_from_string(const sstring&);

replica/compaction_group.hh

@@ -297,17 +297,17 @@ public:
 
     const dht::token_range& token_range() const noexcept;
 
-    size_t memtable_count() const noexcept;
+    size_t memtable_count() const;
 
     const compaction_group_ptr& main_compaction_group() const noexcept;
     const std::vector<compaction_group_ptr>& split_ready_compaction_groups() const;
     compaction_group_ptr& select_compaction_group(locator::tablet_range_side) noexcept;
 
-    uint64_t live_disk_space_used() const noexcept;
+    uint64_t live_disk_space_used() const;
 
-    void for_each_compaction_group(std::function<void(const compaction_group_ptr&)> action) const noexcept;
-    utils::small_vector<compaction_group_ptr, 3> compaction_groups() noexcept;
-    utils::small_vector<const_compaction_group_ptr, 3> compaction_groups() const noexcept;
+    void for_each_compaction_group(std::function<void(const compaction_group_ptr&)> action) const;
+    utils::small_vector<compaction_group_ptr, 3> compaction_groups();
+    utils::small_vector<const_compaction_group_ptr, 3> compaction_groups() const;
 
     utils::small_vector<compaction_group_ptr, 3> split_unready_groups() const;
     bool split_unready_groups_are_empty() const;
@@ -430,7 +430,7 @@ public:
     virtual storage_group& storage_group_for_token(dht::token) const = 0;
     virtual utils::chunked_vector<storage_group_ptr> storage_groups_for_token_range(dht::token_range tr) const = 0;
 
-    virtual locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const noexcept = 0;
+    virtual locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const = 0;
     virtual bool all_storage_groups_split() = 0;
     virtual future<> split_all_storage_groups(tasks::task_info tablet_split_task_info) = 0;
     virtual future<> maybe_split_compaction_group_of(size_t idx) = 0;

replica/database.hh

@@ -1133,7 +1133,7 @@ public:
 
     // The tablet filter is used to not double account migrating tablets, so it's important that
     // only one of pending or leaving replica is accounted based on current migration stage.
-    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const noexcept;
+    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const;
 
     const db::view::stats& get_view_stats() const {
         return _view_stats;

replica/distributed_loader.cc

@@ -234,18 +234,12 @@ distributed_loader::get_sstables_from_upload_dir(sharded<replica::database>& db,
 }
 
 future<std::tuple<table_id, std::vector<std::vector<sstables::shared_sstable>>>>
-distributed_loader::get_sstables_from_object_store(sharded<replica::database>& db, sstring ks, sstring cf, std::vector<sstring> sstables, sstring endpoint, sstring bucket, sstring prefix, sstables::sstable_open_config cfg, std::function<seastar::abort_source*()> get_abort_src) {
-    return get_sstables_from(db, ks, cf, cfg, [bucket, endpoint, prefix, sstables=std::move(sstables), &get_abort_src, &db] (auto& global_table, auto& directory) {
+distributed_loader::get_sstables_from_object_store(sharded<replica::database>& db, sstring ks, sstring cf, std::vector<sstring> sstables, sstring endpoint, sstring type, sstring bucket, sstring prefix, sstables::sstable_open_config cfg, std::function<seastar::abort_source*()> get_abort_src) {
+    return get_sstables_from(db, ks, cf, cfg, [bucket, endpoint, type, prefix, sstables=std::move(sstables), &get_abort_src] (auto& global_table, auto& directory) {
         return directory.start(global_table.as_sharded_parameter(),
-            sharded_parameter([bucket, endpoint, prefix, &get_abort_src, &db] {
-                auto eps = db.local().get_config().object_storage_endpoints() 
-                    | std::views::filter([&endpoint](auto& ep) { return ep.key() == endpoint; })
-                    ;
-                if (eps.empty()) {
-                    throw std::invalid_argument(fmt::format("Undefined endpoint {}", endpoint));
-                }
+            sharded_parameter([bucket, endpoint, type, prefix, &get_abort_src] {
                 seastar::abort_source* as = get_abort_src ? get_abort_src() : nullptr;
-                auto opts = data_dictionary::make_object_storage_options(endpoint, eps.front().type(), bucket, prefix, as);
+                auto opts = data_dictionary::make_object_storage_options(endpoint, type, bucket, prefix, as);
                 return make_lw_shared<const data_dictionary::storage_options>(std::move(opts));
             }),
             sstables,

replica/distributed_loader.hh

@@ -92,7 +92,7 @@ public:
     static future<std::tuple<table_id, std::vector<std::vector<sstables::shared_sstable>>>>
             get_sstables_from_upload_dir(sharded<replica::database>& db, sstring ks, sstring cf, sstables::sstable_open_config cfg);
     static future<std::tuple<table_id, std::vector<std::vector<sstables::shared_sstable>>>>
-            get_sstables_from_object_store(sharded<replica::database>& db, sstring ks, sstring cf, std::vector<sstring> sstables, sstring endpoint, sstring bucket, sstring prefix, sstables::sstable_open_config cfg, std::function<seastar::abort_source*()> = {});
+            get_sstables_from_object_store(sharded<replica::database>& db, sstring ks, sstring cf, std::vector<sstring> sstables, sstring endpoint, sstring type, sstring bucket, sstring prefix, sstables::sstable_open_config cfg, std::function<seastar::abort_source*()> = {});
     static future<> process_upload_dir(sharded<replica::database>& db, sharded<db::view::view_builder>& vb, sharded<db::view::view_building_worker>& vbw, sstring ks_name, sstring cf_name, bool skip_cleanup, bool skip_reshape);
 };
 

replica/table.cc

@@ -708,7 +708,7 @@ public:
         return *_single_sg;
     }
 
-    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)>) const noexcept override {
+    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)>) const override {
         return locator::combined_load_stats{
             .table_ls = locator::table_load_stats{
                             .size_in_bytes = _single_sg->live_disk_space_used(),
@@ -874,7 +874,7 @@ public:
         return storage_group_for_id(storage_group_of(token).first);
     }
 
-    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const noexcept override;
+    locator::combined_load_stats table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const override;
     bool all_storage_groups_split() override;
     future<> split_all_storage_groups(tasks::task_info tablet_split_task_info) override;
     future<> maybe_split_compaction_group_of(size_t idx) override;
@@ -922,7 +922,7 @@ compaction_group_ptr& storage_group::select_compaction_group(locator::tablet_ran
     return _main_cg;
 }
 
-void storage_group::for_each_compaction_group(std::function<void(const compaction_group_ptr&)> action) const noexcept {
+void storage_group::for_each_compaction_group(std::function<void(const compaction_group_ptr&)> action) const {
     action(_main_cg);
     for (auto& cg : _merging_groups) {
         action(cg);
@@ -932,7 +932,7 @@ void storage_group::for_each_compaction_group(std::function<void(const compactio
     }
 }
 
-utils::small_vector<compaction_group_ptr, 3> storage_group::compaction_groups() noexcept {
+utils::small_vector<compaction_group_ptr, 3> storage_group::compaction_groups() {
     utils::small_vector<compaction_group_ptr, 3> cgs;
     for_each_compaction_group([&cgs] (const compaction_group_ptr& cg) {
         cgs.push_back(cg);
@@ -940,7 +940,7 @@ utils::small_vector<compaction_group_ptr, 3> storage_group::compaction_groups()
     return cgs;
 }
 
-utils::small_vector<const_compaction_group_ptr, 3> storage_group::compaction_groups() const noexcept {
+utils::small_vector<const_compaction_group_ptr, 3> storage_group::compaction_groups() const {
     utils::small_vector<const_compaction_group_ptr, 3> cgs;
     for_each_compaction_group([&cgs] (const compaction_group_ptr& cg) {
         cgs.push_back(cg);
@@ -1890,7 +1890,7 @@ sstables::file_size_stats compaction_group::live_disk_space_used_full_stats() co
     return _main_sstables->get_file_size_stats() + _maintenance_sstables->get_file_size_stats();
 }
 
-uint64_t storage_group::live_disk_space_used() const noexcept {
+uint64_t storage_group::live_disk_space_used() const {
     auto cgs = const_cast<storage_group&>(*this).compaction_groups();
     return std::ranges::fold_left(cgs | std::views::transform(std::mem_fn(&compaction_group::live_disk_space_used)), uint64_t(0), std::plus{});
 }
@@ -2813,7 +2813,7 @@ void table::on_flush_timer() {
     });
 }
 
-locator::combined_load_stats tablet_storage_group_manager::table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const noexcept {
+locator::combined_load_stats tablet_storage_group_manager::table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const {
     locator::table_load_stats table_stats;
     table_stats.split_ready_seq_number = _split_ready_seq_number;
 
@@ -2836,7 +2836,7 @@ locator::combined_load_stats tablet_storage_group_manager::table_load_stats(std:
     };
 }
 
-locator::combined_load_stats table::table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const noexcept {
+locator::combined_load_stats table::table_load_stats(std::function<bool(const locator::tablet_map&, locator::global_tablet_id)> tablet_filter) const {
     return _sg_manager->table_load_stats(std::move(tablet_filter));
 }
 
@@ -3453,7 +3453,7 @@ size_t compaction_group::memtable_count() const noexcept {
     return _memtables->size();
 }
 
-size_t storage_group::memtable_count() const noexcept {
+size_t storage_group::memtable_count() const {
     return std::ranges::fold_left(compaction_groups() | std::views::transform(std::mem_fn(&compaction_group::memtable_count)), size_t(0), std::plus{});
 }
 

scripts/pull_github_pr.sh

@@ -38,8 +38,9 @@ for required in jq curl; do
 	fi
 done
 
-FORCE=0
 ALLOW_SUBMODULE=0
+ALLOW_UNSTABLE=0
+ALLOW_ANY_BRANCH=0
 
 function print_usage {
 cat << EOF
@@ -60,12 +61,18 @@ Options:
 -h
     Print this help message and exit.
 
---force
-    Do not check current branch to be next*
-    Do not check jenkins job status
-
 --allow-submodule
     Allow a PR to update a submudule
+
+--allow-unstable
+    Do not check jenkins job status
+
+--allow-any-branch
+    Merge PR even if target branch is not next
+
+--force
+    Sets all above --allow-* options
+
 EOF
 }
 
@@ -73,13 +80,23 @@ while [[ $# -gt 0 ]]
 do
     case $1 in
         "--force"|"-f")
-            FORCE=1
+            ALLOW_UNSTABLE=1
+            ALLOW_SUBMODULE=1
+            ALLOW_ANY_BRANCH=1
             shift 1
             ;;
         --allow-submodule)
             ALLOW_SUBMODULE=1
             shift
             ;;
+        --allow-unstable)
+            ALLOW_UNSTABLE=1
+            shift
+            ;;
+        --allow-any-branch)
+            ALLOW_ANY_BRANCH=1
+            shift
+            ;;
         +([0-9]))
             PR_NUM=$1
             shift 1
@@ -147,7 +164,7 @@ check_jenkins_job_status() {
   fi
 }
 
-if [[ $FORCE -eq 0 ]]; then
+if [[ $ALLOW_UNSTABLE -eq 0 ]]; then
   check_jenkins_job_status
 fi
 
@@ -179,17 +196,19 @@ echo -n "Fetching full name of author $PR_LOGIN... "
 USER_NAME=$(curl -s "https://api.github.com/users/$PR_LOGIN" | jq -r .name)
 echo "$USER_NAME"
 
-BASE_BRANCH=$(jq -r .base.ref <<< $PR_DATA)
-CURRENT_BRANCH=$(git rev-parse --abbrev-ref HEAD)
-TARGET_BASE="unknown"
-if [[ ${BASE_BRANCH} == master ]]; then
-    TARGET_BASE="next"
-elif [[ ${BASE_BRANCH}  == branch-* ]]; then
-    TARGET_BASE=${BASE_BRANCH//branch/next}
-fi
-if [[ "${CURRENT_BRANCH}" != "${TARGET_BASE}" ]]; then
-    echo "Merging into wrong next, want ${TARGET_BASE}, have ${CURRENT_BRANCH}"
-    exit 1
+if [[ $ALLOW_ANY_BRANCH -eq 0 ]]; then
+    BASE_BRANCH=$(jq -r .base.ref <<< $PR_DATA)
+    CURRENT_BRANCH=$(git rev-parse --abbrev-ref HEAD)
+    TARGET_BASE="unknown"
+    if [[ ${BASE_BRANCH} == master ]]; then
+        TARGET_BASE="next"
+    elif [[ ${BASE_BRANCH}  == branch-* ]]; then
+        TARGET_BASE=${BASE_BRANCH//branch/next}
+    fi
+    if [[ "${CURRENT_BRANCH}" != "${TARGET_BASE}" ]]; then
+        echo "Merging into wrong next, want ${TARGET_BASE}, have ${CURRENT_BRANCH}. Use --allow-any-branch or --force to skip this check"
+        exit 1
+    fi
 fi
 
 git fetch "$REMOTE" pull/$PR_NUM/head

sstables/sstables.cc

@@ -2485,11 +2485,6 @@ void sstable::validate_originating_host_id() const {
         }
         return;
     }
-
-    if (*originating_host_id != local_host_id) {
-        // FIXME refrain from throwing an exception because of #10148
-        sstlog.warn("Host id {} does not match local host id {} while validating SSTable: {}. Load foreign SSTables via the upload dir instead.", *originating_host_id, local_host_id, get_filename());
-    }
 }
 
 sstring sstable::component_basename(const sstring& ks, const sstring& cf, version_types version, generation_type generation,

sstables/sstables_manager.cc

@@ -135,13 +135,17 @@ future<> storage_manager::update_config(const db::config& cfg) {
     co_return;
 }
 
-shared_ptr<sstables::object_storage_client> storage_manager::get_endpoint_client(sstring endpoint) {
+auto storage_manager::get_endpoint(const sstring& endpoint) -> object_storage_endpoint& {
     auto found = _object_storage_endpoints.find(endpoint);
     if (found == _object_storage_endpoints.end()) {
         smlogger.error("unable to find {} in configured object-storage endpoints", endpoint);
         throw std::invalid_argument(format("endpoint {} not found", endpoint));
     }
-    auto& ep = found->second;
+    return found->second;
+}
+
+shared_ptr<sstables::object_storage_client> storage_manager::get_endpoint_client(sstring endpoint) {
+    auto& ep = get_endpoint(endpoint);
     if (ep.client == nullptr) {
         ep.client = make_object_storage_client(ep.cfg, _object_storage_clients_memory, [&ct = container()] (std::string ep) {
             return ct.local().get_endpoint_client(ep);
@@ -150,6 +154,10 @@ shared_ptr<sstables::object_storage_client> storage_manager::get_endpoint_client
     return ep.client;
 }
 
+sstring storage_manager::get_endpoint_type(sstring endpoint) {
+    return get_endpoint(endpoint).cfg.type();
+}
+
 bool storage_manager::is_known_endpoint(sstring endpoint) const {
     return _object_storage_endpoints.contains(endpoint);
 }

sstables/sstables_manager.hh

@@ -70,6 +70,7 @@ class storage_manager : public peering_sharded_service<storage_manager> {
     seastar::metrics::metric_groups metrics;
 
     future<> update_config(const db::config&);
+    object_storage_endpoint& get_endpoint(const sstring& ep);
 
 public:
     struct config {
@@ -80,6 +81,7 @@ public:
     storage_manager(const db::config&, config cfg);
     shared_ptr<object_storage_client> get_endpoint_client(sstring endpoint);
     bool is_known_endpoint(sstring endpoint) const;
+    sstring get_endpoint_type(sstring endpoint);
     future<> stop();
     std::vector<sstring> endpoints(sstring type = "") const noexcept;
 };

sstables_loader.cc

@@ -205,6 +205,13 @@ private:
     }
 
     bool tablet_in_scope(locator::tablet_id) const;
+
+    friend future<std::vector<tablet_sstable_collection>> get_sstables_for_tablets_for_tests(const std::vector<sstables::shared_sstable>& sstables,
+                                                                                             std::vector<dht::token_range>&& tablets_ranges);
+    // Pay attention, while working with tablet ranges, the `erm` must be held alive as long as we retrieve (and use here) tablet ranges from
+    // the tablet map. This is already done when using `tablet_sstable_streamer` class but tread carefully if you plan to use this method somewhere else.
+    static future<std::vector<tablet_sstable_collection>> get_sstables_for_tablets(const std::vector<sstables::shared_sstable>& sstables,
+                                                                                   std::vector<dht::token_range>&& tablets_ranges);
 };
 
 host_id_vector_replica_set sstable_streamer::get_endpoints(const dht::token& token) const {
@@ -343,55 +350,52 @@ public:
     }
 };
 
+future<std::vector<tablet_sstable_collection>> tablet_sstable_streamer::get_sstables_for_tablets(const std::vector<sstables::shared_sstable>& sstables,
+                                                                                                 std::vector<dht::token_range>&& tablets_ranges) {
+    auto tablets_sstables =
+        tablets_ranges | std::views::transform([](auto range) { return tablet_sstable_collection{.tablet_range = range}; }) | std::ranges::to<std::vector>();
+    if (sstables.empty() || tablets_sstables.empty()) {
+        co_return std::move(tablets_sstables);
+    }
+    // sstables are sorted by first key in reverse order.
+    auto reversed_sstables = sstables | std::views::reverse;
+
+    for (auto& [tablet_range, sstables_fully_contained, sstables_partially_contained] : tablets_sstables) {
+        for (const auto& sst : reversed_sstables) {
+            auto sst_first = sst->get_first_decorated_key().token();
+            auto sst_last = sst->get_last_decorated_key().token();
+
+            // SSTable entirely after tablet -> no further SSTables (larger keys) can overlap
+            if (tablet_range.after(sst_first, dht::token_comparator{})) {
+                break;
+            }
+            // SSTable entirely before tablet -> skip and continue scanning later (larger keys)
+            if (tablet_range.before(sst_last, dht::token_comparator{})) {
+                continue;
+            }
+
+            if (tablet_range.contains(dht::token_range{sst_first, sst_last}, dht::token_comparator{})) {
+                sstables_fully_contained.push_back(sst);
+            } else {
+                sstables_partially_contained.push_back(sst);
+            }
+            co_await coroutine::maybe_yield();
+        }
+    }
+    co_return std::move(tablets_sstables);
+}
+
 future<> tablet_sstable_streamer::stream(shared_ptr<stream_progress> progress) {
     if (progress) {
         progress->start(_tablet_map.tablet_count());
     }
 
-    // sstables are sorted by first key in reverse order.
-    auto sstable_it = _sstables.rbegin();
-
-    for (auto tablet_id : _tablet_map.tablet_ids() | std::views::filter([this] (auto tid) { return tablet_in_scope(tid); })) {
-        auto tablet_range = _tablet_map.get_token_range(tablet_id);
-
-        auto sstable_token_range = [] (const sstables::shared_sstable& sst) {
-            return dht::token_range(sst->get_first_decorated_key().token(),
-                                    sst->get_last_decorated_key().token());
-        };
-
-        std::vector<sstables::shared_sstable> sstables_fully_contained;
-        std::vector<sstables::shared_sstable> sstables_partially_contained;
-
-        // sstable is exhausted if its last key is before the current tablet range
-        auto exhausted = [&tablet_range] (const sstables::shared_sstable& sst) {
-            return tablet_range.before(sst->get_last_decorated_key().token(), dht::token_comparator{});
-        };
-        while (sstable_it != _sstables.rend() && exhausted(*sstable_it)) {
-            sstable_it++;
-        }
-
-        for (auto sst_it = sstable_it; sst_it != _sstables.rend(); sst_it++) {
-            auto sst_token_range = sstable_token_range(*sst_it);
-
-            // sstables are sorted by first key, so should skip this SSTable since it
-            // doesn't overlap with the current tablet range.
-            if (!tablet_range.overlaps(sst_token_range, dht::token_comparator{})) {
-                // If the start of the next SSTable's token range lies beyond the current tablet's token
-                // range, we can safely conclude that no more relevant SSTables remain for this tablet.
-                if (tablet_range.after(sst_token_range.start()->value(), dht::token_comparator{})) {
-                    break;
-                }
-                continue;
-            }
-
-            if (tablet_range.contains(sst_token_range, dht::token_comparator{})) {
-                sstables_fully_contained.push_back(*sst_it);
-            } else {
-                sstables_partially_contained.push_back(*sst_it);
-            }
-            co_await coroutine::maybe_yield();
-        }
+    auto classified_sstables = co_await get_sstables_for_tablets(
+        _sstables, _tablet_map.tablet_ids() | std::views::filter([this](auto tid) { return tablet_in_scope(tid); }) | std::views::transform([this](auto tid) {
+                       return _tablet_map.get_token_range(tid);
+                   }) | std::ranges::to<std::vector>());
 
+    for (auto& [tablet_range, sstables_fully_contained, sstables_partially_contained] : classified_sstables) {
         auto per_tablet_progress = make_shared<per_tablet_stream_progress>(
             progress,
             sstables_fully_contained.size() + sstables_partially_contained.size());
@@ -751,8 +755,9 @@ future<> sstables_loader::download_task_impl::run() {
     };
     llog.debug("Loading sstables from {}({}/{})", _endpoint, _bucket, _prefix);
 
+    auto ep_type = _loader.local()._storage_manager.get_endpoint_type(_endpoint);
     std::vector<seastar::abort_source> shard_aborts(smp::count);
-    auto [ table_id, sstables_on_shards ] = co_await replica::distributed_loader::get_sstables_from_object_store(_loader.local()._db, _ks, _cf, _sstables, _endpoint, _bucket, _prefix, cfg, [&] {
+    auto [ table_id, sstables_on_shards ] = co_await replica::distributed_loader::get_sstables_from_object_store(_loader.local()._db, _ks, _cf, _sstables, _endpoint, ep_type, _bucket, _prefix, cfg, [&] {
         return &shard_aborts[this_shard_id()];
     });
     llog.debug("Streaming sstables from {}({}/{})", _endpoint, _bucket, _prefix);
@@ -832,3 +837,7 @@ future<tasks::task_id> sstables_loader::download_new_sstables(sstring ks_name, s
                                                                                        std::move(prefix), std::move(sstables), scope, primary_replica_only(primary_replica));
     co_return task->id();
 }
+future<std::vector<tablet_sstable_collection>> get_sstables_for_tablets_for_tests(const std::vector<sstables::shared_sstable>& sstables,
+                                                                                  std::vector<dht::token_range>&& tablets_ranges) {
+    return tablet_sstable_streamer::get_sstables_for_tablets(sstables, std::move(tablets_ranges));
+}

sstables_loader.hh

@@ -10,6 +10,8 @@
 
 #include <vector>
 #include <seastar/core/sharded.hh>
+#include "dht/i_partitioner_fwd.hh"
+#include "dht/token.hh"
 #include "schema/schema_fwd.hh"
 #include "sstables/shared_sstable.hh"
 #include "tasks/task_manager.hh"
@@ -152,3 +154,18 @@ struct fmt::formatter<sstables_loader::stream_scope> : fmt::formatter<string_vie
         }
     }
 };
+
+struct tablet_sstable_collection {
+    dht::token_range tablet_range;
+    std::vector<sstables::shared_sstable> sstables_fully_contained;
+    std::vector<sstables::shared_sstable> sstables_partially_contained;
+};
+
+// This function is intended for test purposes only.
+// It assigns the given sstables to the given tablet ranges based on token containment.
+// It returns a vector of tablet_sstable_collection, each containing the tablet range
+// and the sstables that are fully or partially contained within that range.
+// The prerequisite is the tablet ranges are sorted by the range in ascending order and non-overlapping.
+// Another prerequisite is that the sstables' token ranges are sorted by its `start` in descending order.
+future<std::vector<tablet_sstable_collection>> get_sstables_for_tablets_for_tests(const std::vector<sstables::shared_sstable>& sstables,
+                                                                                  std::vector<dht::token_range>&& tablets_ranges);

test/boost/CMakeLists.txt

@@ -370,6 +370,7 @@ add_scylla_test(combined_tests
     sstable_compression_config_test.cc
     sstable_directory_test.cc
     sstable_set_test.cc
+    sstable_tablet_streaming.cc
     statement_restrictions_test.cc
     storage_proxy_test.cc
     tablets_test.cc

test/boost/network_topology_strategy_test.cc

@@ -1450,8 +1450,7 @@ SEASTAR_THREAD_TEST_CASE(tablets_simple_rack_aware_view_pairing_test) {
         std::map<sstring, replication_strategy_config_option> options;
         for (const auto& dc : option_dcs) {
             auto num_racks = node_count_per_rack.at(dc).size();
-            auto max_rf_factor = std::ranges::min(std::ranges::views::transform(node_count_per_rack.at(dc), [] (auto& x) { return x.second; }));
-            auto rf = num_racks * tests::random::get_int(1UL, max_rf_factor);
+            auto rf = num_racks;
             options.emplace(dc, fmt::to_string(rf));
         }
         return options;
@@ -1487,8 +1486,7 @@ SEASTAR_THREAD_TEST_CASE(tablets_simple_rack_aware_view_pairing_test) {
     // Test tablets rack-aware base-view pairing
     auto base_token = dht::token::get_random_token();
     auto view_token = dht::token::get_random_token();
-    bool use_legacy_self_pairing = false;
-    bool use_tablets_basic_rack_aware_view_pairing = true;
+    bool use_tablets = true;
     const auto& base_replicas = base_tmap.get_tablet_info(base_tmap.get_tablet_id(base_token)).replicas;
     replica::cf_stats cf_stats;
     std::unordered_map<locator::host_id, locator::host_id> base_to_view_pairing;
@@ -1502,8 +1500,7 @@ SEASTAR_THREAD_TEST_CASE(tablets_simple_rack_aware_view_pairing_test) {
             *ars_ptr,
             base_token,
             view_token,
-            use_legacy_self_pairing,
-            use_tablets_basic_rack_aware_view_pairing,
+            use_tablets,
             cf_stats).natural_endpoint;
 
         // view pair must be found
@@ -1525,181 +1522,6 @@ SEASTAR_THREAD_TEST_CASE(tablets_simple_rack_aware_view_pairing_test) {
     }
 }
 
-// Called in a seastar thread
-void test_complex_rack_aware_view_pairing_test(bool more_or_less) {
-    auto my_address = gms::inet_address("localhost");
-
-    // Create the RackInferringSnitch
-    snitch_config cfg;
-    cfg.listen_address = my_address;
-    cfg.broadcast_address = my_address;
-    cfg.name = "RackInferringSnitch";
-    sharded<snitch_ptr> snitch;
-    snitch.start(cfg).get();
-    auto stop_snitch = defer([&snitch] { snitch.stop().get(); });
-    snitch.invoke_on_all(&snitch_ptr::start).get();
-
-    locator::token_metadata::config tm_cfg;
-    tm_cfg.topo_cfg.this_endpoint = my_address;
-    tm_cfg.topo_cfg.local_dc_rack = { snitch.local()->get_datacenter(), snitch.local()->get_rack() };
-
-    std::map<sstring, size_t> node_count_per_dc;
-    std::map<sstring, std::map<sstring, size_t>> node_count_per_rack;
-    std::vector<ring_point> ring_points;
-
-    auto& random_engine = seastar::testing::local_random_engine;
-    unsigned shard_count = 2;
-    size_t num_dcs = 1 + tests::random::get_int(3);
-
-    // Generate a random cluster
-    double point = 1;
-    for (size_t dc = 0; dc < num_dcs; ++dc) {
-        sstring dc_name = fmt::format("{}", 100 + dc);
-        size_t num_racks = 2 + tests::random::get_int(4);
-        for (size_t rack = 0; rack < num_racks; ++rack) {
-            sstring rack_name = fmt::format("{}", 10 + rack);
-            size_t rack_nodes = 1 + tests::random::get_int(2);
-            for (size_t i = 1; i <= rack_nodes; ++i) {
-                ring_points.emplace_back(point, inet_address(format("192.{}.{}.{}", dc_name, rack_name, i)));
-                node_count_per_dc[dc_name]++;
-                node_count_per_rack[dc_name][rack_name]++;
-                point++;
-            }
-        }
-    }
-
-    testlog.debug("node_count_per_rack={}", node_count_per_rack);
-
-    // Initialize the token_metadata
-    locator::shared_token_metadata stm([] () noexcept { return db::schema_tables::hold_merge_lock(); }, tm_cfg);
-    auto stop_stm = deferred_stop(stm);
-    stm.mutate_token_metadata([&] (token_metadata& tm) -> future<> {
-        auto& topo = tm.get_topology();
-        for (const auto& [ring_point, endpoint, id] : ring_points) {
-            std::unordered_set<token> tokens;
-            tokens.insert(token{tests::d2t(ring_point / ring_points.size())});
-            topo.add_node(id, make_endpoint_dc_rack(endpoint), locator::node::state::normal, shard_count);
-            co_await tm.update_normal_tokens(std::move(tokens), id);
-        }
-    }).get();
-
-    auto base_schema = schema_builder("ks", "base")
-        .with_column("k", utf8_type, column_kind::partition_key)
-        .with_column("v", utf8_type)
-        .build();
-
-    auto view_schema = schema_builder("ks", "view")
-        .with_column("v", utf8_type, column_kind::partition_key)
-        .with_column("k", utf8_type)
-        .build();
-
-    auto tmptr = stm.get();
-
-    // Create the replication strategy
-    auto make_random_options = [&] () {
-        auto option_dcs = node_count_per_dc | std::views::keys | std::ranges::to<std::vector>();
-        std::shuffle(option_dcs.begin(), option_dcs.end(), random_engine);
-        std::map<sstring, replication_strategy_config_option> options;
-        for (const auto& dc : option_dcs) {
-            auto num_racks = node_count_per_rack.at(dc).size();
-            auto rf = more_or_less ?
-                    tests::random::get_int(num_racks, node_count_per_dc[dc]) :
-                    tests::random::get_int(1UL, num_racks);
-            options.emplace(dc, fmt::to_string(rf));
-        }
-        return options;
-    };
-
-    auto options = make_random_options();
-    size_t tablet_count = 1 + tests::random::get_int(99);
-    testlog.debug("tablet_count={} rf_options={}", tablet_count, options);
-    locator::replication_strategy_params params(options, tablet_count, std::nullopt);
-    auto ars_ptr = abstract_replication_strategy::create_replication_strategy(
-            "NetworkTopologyStrategy", params, tmptr->get_topology());
-    auto tab_awr_ptr = ars_ptr->maybe_as_tablet_aware();
-    BOOST_REQUIRE(tab_awr_ptr);
-    auto base_tmap = tab_awr_ptr->allocate_tablets_for_new_table(base_schema, tmptr, 1).get();
-    auto base_table_id = base_schema->id();
-    testlog.debug("base_table_id={}", base_table_id);
-    auto view_table_id = view_schema->id();
-    auto view_tmap = tab_awr_ptr->allocate_tablets_for_new_table(view_schema, tmptr, 1).get();
-    testlog.debug("view_table_id={}", view_table_id);
-
-    stm.mutate_token_metadata([&] (token_metadata& tm) -> future<> {
-        tm.tablets().set_tablet_map(base_table_id, co_await base_tmap.clone_gently());
-        tm.tablets().set_tablet_map(view_table_id, co_await view_tmap.clone_gently());
-    }).get();
-
-    tmptr = stm.get();
-    auto base_erm = tab_awr_ptr->make_replication_map(base_table_id, tmptr);
-    auto view_erm = tab_awr_ptr->make_replication_map(view_table_id, tmptr);
-
-    auto& topology = tmptr->get_topology();
-    testlog.debug("topology: {}", topology.get_datacenter_racks());
-
-    // Test tablets rack-aware base-view pairing
-    auto base_token = dht::token::get_random_token();
-    auto view_token = dht::token::get_random_token();
-    bool use_legacy_self_pairing = false;
-    bool use_tablets_basic_rack_aware_view_pairing = true;
-    const auto& base_replicas = base_tmap.get_tablet_info(base_tmap.get_tablet_id(base_token)).replicas;
-    replica::cf_stats cf_stats;
-    std::unordered_map<locator::host_id, locator::host_id> base_to_view_pairing;
-    std::unordered_map<locator::host_id, locator::host_id> view_to_base_pairing;
-    std::unordered_map<sstring, size_t> same_rack_pairs;
-    std::unordered_map<sstring, size_t> cross_rack_pairs;
-    for (const auto& base_replica : base_replicas) {
-        auto& base_host = base_replica.host;
-        auto view_ep_opt = db::view::get_view_natural_endpoint(
-            base_host,
-            base_erm,
-            view_erm,
-            *ars_ptr,
-            base_token,
-            view_token,
-            use_legacy_self_pairing,
-            use_tablets_basic_rack_aware_view_pairing,
-            cf_stats).natural_endpoint;
-
-        // view pair must be found
-        if (!view_ep_opt) {
-            BOOST_FAIL(format("Could not pair base_host={} base_token={} view_token={}", base_host, base_token, view_token));
-        }
-        BOOST_REQUIRE(view_ep_opt);
-        auto& view_ep = *view_ep_opt;
-
-        // Assert pairing uniqueness
-        auto [base_it, inserted_base_pair] = base_to_view_pairing.emplace(base_host, view_ep);
-        BOOST_REQUIRE(inserted_base_pair);
-        auto [view_it, inserted_view_pair] = view_to_base_pairing.emplace(view_ep, base_host);
-        BOOST_REQUIRE(inserted_view_pair);
-
-        auto& base_location = topology.find_node(base_host)->dc_rack();
-        auto& view_location = topology.find_node(view_ep)->dc_rack();
-
-        // Assert dc- and rack- aware pairing
-        BOOST_REQUIRE_EQUAL(base_location.dc, view_location.dc);
-
-        if (base_location.rack == view_location.rack) {
-            same_rack_pairs[base_location.dc]++;
-        } else {
-            cross_rack_pairs[base_location.dc]++;
-        }
-    }
-    for (const auto& [dc, rf_opt] : options) {
-        auto rf = locator::get_replication_factor(rf_opt);
-        BOOST_REQUIRE_EQUAL(same_rack_pairs[dc] + cross_rack_pairs[dc], rf);
-    }
-}
-
-SEASTAR_THREAD_TEST_CASE(tablets_complex_rack_aware_view_pairing_test_rf_lt_racks) {
-    test_complex_rack_aware_view_pairing_test(false);
-}
-
-SEASTAR_THREAD_TEST_CASE(tablets_complex_rack_aware_view_pairing_test_rf_gt_racks) {
-    test_complex_rack_aware_view_pairing_test(true);
-}
-
 SEASTAR_THREAD_TEST_CASE(test_rack_diff) {
     BOOST_REQUIRE(diff_racks({}, {}).empty());
 

test/boost/sstable_tablet_streaming.cc

@@ -0,0 +1,367 @@
+/*
+ * Copyright (C) 2025-present ScyllaDB
+ */
+
+/*
+ * SPDX-License-Identifier: LicenseRef-ScyllaDB-Source-Available-1.0
+ */
+
+#undef SEASTAR_TESTING_MAIN
+#include <seastar/testing/test_case.hh>
+#include "dht/token.hh"
+#include "sstable_test.hh"
+#include "sstables_loader.hh"
+#include "test/lib/sstable_test_env.hh"
+
+BOOST_AUTO_TEST_SUITE(sstable_tablet_streaming_test)
+
+using namespace sstables;
+
+std::vector<shared_sstable> make_sstables_with_ranges(test_env& env, const std::vector<std::pair<int64_t, int64_t>>& ranges) {
+    std::vector<shared_sstable> ssts;
+    for (const auto& [first, last] : ranges) {
+        auto sst = env.make_sstable(uncompressed_schema(), uncompressed_dir());
+        test(sst).set_first_and_last_keys(dht::decorated_key(dht::token{first}, partition_key(std::vector<bytes>{"1"})),
+                                          dht::decorated_key(dht::token{last}, partition_key(std::vector<bytes>{"1"})));
+        ssts.push_back(std::move(sst));
+    }
+    // By sorting SSTables by their primary key, we enable runs to be
+    // streamed incrementally. Overlapping fragments can be deduplicated,
+    // reducing the amount of data sent over the wire. Elements are
+    // popped from the back of the vector, so we sort in descending
+    // order to begin with the smaller tokens.
+    // See sstable_streamer constructor for more details.
+    std::ranges::sort(ssts, [](const shared_sstable& x, const shared_sstable& y) { return x->compare_by_first_key(*y) > 0; });
+    return ssts;
+}
+
+std::vector<dht::token_range> get_tablet_sstable_collection(auto&&... tablet_ranges) {
+    // tablet ranges are left-non-inclusive, see `tablet_map::get_token_range` for details
+    std::vector<dht::token_range> collections{dht::token_range::make({tablet_ranges.start()->value(), false}, {tablet_ranges.end()->value(), true})...};
+
+    std::sort(collections.begin(), collections.end(), [](auto const& a, auto const& b) { return a.start()->value() < b.start()->value(); });
+
+    return collections;
+}
+
+#define REQUIRE_WITH_CONTEXT(sstables, expected_size)                                                                                                          \
+    BOOST_TEST_CONTEXT("Testing with ranges: " << [&] {                                                                                                        \
+        std::stringstream ss;                                                                                                                                  \
+        for (const auto& sst : (sstables)) {                                                                                                                   \
+            ss << dht::token_range(sst->get_first_decorated_key().token(), sst->get_last_decorated_key().token()) << ", ";                                     \
+        }                                                                                                                                                      \
+        return ss.str();                                                                                                                                       \
+    }())                                                                                                                                                       \
+    BOOST_REQUIRE_EQUAL(sstables.size(), expected_size)
+
+SEASTAR_TEST_CASE(test_streaming_ranges_distribution) {
+    return test_env::do_with_async([](test_env& env) {
+        // 1) Exact boundary equality: SSTable == tablet
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {5, 10},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+        }
+
+        // 2) Single-point overlaps at start/end
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {4, 5},   // touches start, non-inclusive, skip
+                                                      {10, 11}, // touches end
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+        }
+
+        // 3) Tablet fully inside a large SSTable
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 20},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+        }
+
+        // 4) Multiple SSTables fully contained in tablet
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {6, 7},
+                                                      {7, 8},
+                                                      {8, 9},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 3);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 0);
+        }
+
+        // 5) Two overlapping but not fully contained SSTables
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 6},  // overlaps at left
+                                                      {9, 15}, // overlaps at right
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 2);
+        }
+
+        // 6) Unsorted input (helper sorts) + mixed overlaps
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{50}, dht::token{100}});
+            // Intentionally unsorted by first token
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {120, 130},
+                                                      {0, 10},
+                                                      {60, 70},  // fully contained
+                                                      {40, 55},  // partial
+                                                      {95, 105}, // partial
+                                                      {80, 90},  // fully contained
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 2);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 2);
+        }
+
+        // 7) Empty SSTable list
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            std::vector<shared_sstable> ssts;
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 0);
+        }
+
+        // 8) Tablet outside all SSTables
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{100}, dht::token{200}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {1, 2},
+                                                      {3, 4},
+                                                      {10, 20},
+                                                      {300, 400},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 0);
+        }
+
+        // 9) Boundary adjacency with multiple fragments
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{100}, dht::token{200}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {50, 100},  // touches start -> non-inclusive, skip
+                                                      {100, 120}, // starts at start -> partially contained
+                                                      {180, 200}, // ends at end   -> fully contained
+                                                      {200, 220}, // touches end   -> partial
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 1);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 2);
+        }
+
+        // 10) Large SSTable set where early break should occur
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{1000}, dht::token{2000}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {100, 200},
+                                                      {300, 400},
+                                                      {900, 950},
+                                                      {1001, 1100}, // fully contained
+                                                      {1500, 1600}, // fully contained
+                                                      {2101, 2200}, // entirely after -> should trigger early break in ascending scan
+                                                      {1999, 2100}, // overlap, partially contained
+                                                      {3000, 3100},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 2);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+        }
+
+        // 10) https://github.com/scylladb/scylladb/pull/26980 example, tested
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{4}, dht::token{5}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 5},
+                                                      {0, 3},
+                                                      {2, 5},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            // None fully contained; three partial overlaps
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 2);
+        }
+    });
+}
+
+SEASTAR_TEST_CASE(test_streaming_ranges_distribution_in_tablets) {
+    return test_env::do_with_async([](test_env& env) {
+        {
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}}, dht::token_range{dht::token{11}, dht::token{15}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {5, 10},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 0);
+        }
+
+        {
+            // Multiple tablets with a hole between [10,11]
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{0}, dht::token{4}},
+                                                            dht::token_range{dht::token{5}, dht::token{9}},
+                                                            dht::token_range{dht::token{12}, dht::token{15}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 4},   // T.start==S.start, but non-inclusive -> partial
+                                                      {5, 9},   // same as above
+                                                      {6, 8},   // fully in second tablet
+                                                      {10, 11}, // falls in the hole, should be rejected
+                                                      {8, 13},  // overlaps second and third tablets (partial in both)
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+
+            REQUIRE_WITH_CONTEXT(res[1].sstables_fully_contained, 1);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 2);
+
+            REQUIRE_WITH_CONTEXT(res[2].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[2].sstables_partially_contained, 1);
+        }
+
+        {
+            // SSTables outside any tablet range
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{20}, dht::token{25}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 5},   // before
+                                                      {30, 35}, // after
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 0);
+        }
+
+        {
+            // Edge case: SSTable touching tablet boundary
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{5}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {4, 5},   // touches start, non-inclusive, skip
+                                                      {10, 11}, // touches end
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+        }
+
+        {
+            // No tablets, but some SSTables
+            auto collection = get_tablet_sstable_collection();
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 5},
+                                                      {10, 15},
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            BOOST_REQUIRE_EQUAL(res.size(), 0); // no tablets → nothing to classify
+        }
+
+        {
+            // No SSTables, but some tablets
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{0}, dht::token{5}}, dht::token_range{dht::token{10}, dht::token{15}});
+            std::vector<shared_sstable> ssts; // empty
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 0);
+        }
+
+        {
+            // No tablets and no SSTables
+            auto collection = get_tablet_sstable_collection();
+            std::vector<shared_sstable> ssts; // empty
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+            BOOST_REQUIRE_EQUAL(res.size(), 0);
+        }
+        {
+            // SSTable spanning two tablets
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{0}, dht::token{4}}, dht::token_range{dht::token{5}, dht::token{9}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {2, 7}, // spans both tablets
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            // Tablet [0,4] sees partial overlap
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+
+            // Tablet [5,9] sees partial overlap
+            REQUIRE_WITH_CONTEXT(res[1].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 1);
+        }
+
+        {
+            // SSTable spanning three tablets with a hole in between
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{0}, dht::token{3}},
+                                                            dht::token_range{dht::token{4}, dht::token{6}},
+                                                            dht::token_range{dht::token{8}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {2, 9}, // spans across tablets 1,2,3 and hole [7]
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 1);
+            REQUIRE_WITH_CONTEXT(res[2].sstables_partially_contained, 1);
+        }
+
+        {
+            // SSTable fully covering one tablet and partially overlapping another
+            auto collection = get_tablet_sstable_collection(dht::token_range{dht::token{0}, dht::token{5}}, dht::token_range{dht::token{6}, dht::token{10}});
+            auto ssts = make_sstables_with_ranges(env,
+                                                  {
+                                                      {0, 7}, // fully covers first tablet, partial in second
+                                                  });
+            auto res = get_sstables_for_tablets_for_tests(ssts, std::move(collection)).get();
+
+            REQUIRE_WITH_CONTEXT(res[0].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[0].sstables_partially_contained, 1);
+
+            REQUIRE_WITH_CONTEXT(res[1].sstables_fully_contained, 0);
+            REQUIRE_WITH_CONTEXT(res[1].sstables_partially_contained, 1);
+        }
+    });
+}
+
+BOOST_AUTO_TEST_SUITE_END()

test/boost/view_build_test.cc

@@ -395,7 +395,7 @@ SEASTAR_TEST_CASE(test_builder_with_concurrent_drop) {
             assert_that(msg).is_rows().is_empty();
             msg = e.execute_cql("select * from system_distributed.view_build_status").get();
             assert_that(msg).is_rows().is_empty();
-        });
+        }, 30);
     });
 }
 

test/cluster/dtest/audit_test.py

@@ -25,7 +25,7 @@ from typing import Any, Optional, override
 import pytest
 import requests
 from cassandra import AlreadyExists, AuthenticationFailed, ConsistencyLevel, InvalidRequest, Unauthorized, Unavailable, WriteFailure
-from cassandra.cluster import NoHostAvailable, Session
+from cassandra.cluster import NoHostAvailable, Session, EXEC_PROFILE_DEFAULT
 from cassandra.query import SimpleStatement, named_tuple_factory
 from ccmlib.scylla_node import ScyllaNode, NodeError
 
@@ -1135,6 +1135,14 @@ class TestCQLAudit(AuditTester):
 
             session.execute("DROP TABLE test1")
 
+    def _get_attempt_count(self, session: Session, *, execution_profile=EXEC_PROFILE_DEFAULT, consistency_level: ConsistencyLevel = ConsistencyLevel.ONE) -> int:
+        # dtest env is using FlakyRetryPolicy which has `max_retries` attribute
+        cl_profile = session.execution_profile_clone_update(execution_profile, consistency_level=consistency_level)
+        policy = cl_profile.retry_policy
+        retries = getattr(policy, "max_retries", None)
+        assert retries is not None
+        return 1 + retries
+
     def _test_insert_failure_doesnt_report_success_assign_nodes(self, session: Session = None):
         all_nodes: set[ScyllaNode] = set(self.cluster.nodelist())
         assert len(all_nodes) == 7
@@ -1154,6 +1162,7 @@ class TestCQLAudit(AuditTester):
         for i in range(256):
             stmt = SimpleStatement(f"INSERT INTO ks.test1 (k, v1) VALUES ({i}, 1337)", consistency_level=ConsistencyLevel.THREE)
             session.execute(stmt)
+            attempt_count = self._get_attempt_count(session, consistency_level=ConsistencyLevel.THREE)
 
             token = rows_to_list(session.execute(f"SELECT token(k) FROM ks.test1 WHERE k = {i}"))[0][0]
 
@@ -1168,9 +1177,9 @@ class TestCQLAudit(AuditTester):
                     audit_partition_nodes = [address_to_node[address] for address in audit_nodes]
                     insert_node = address_to_node[insert_node.pop()]
                     kill_node = address_to_node[partitions.pop()]
-                    return audit_partition_nodes, insert_node, kill_node, stmt.query_string
+                    return audit_partition_nodes, insert_node, kill_node, stmt.query_string, attempt_count
 
-        return [], [], None, None
+        return [], [], None, None, None
 
     @pytest.mark.exclude_errors("audit - Unexpected exception when writing log with: node_ip")
     def test_insert_failure_doesnt_report_success(self):
@@ -1192,7 +1201,7 @@ class TestCQLAudit(AuditTester):
             with self.assert_exactly_n_audit_entries_were_added(session, 1):
                 conn.execute(stmt)
 
-        audit_paritition_nodes, insert_node, node_to_stop, query_to_fail = self._test_insert_failure_doesnt_report_success_assign_nodes(session=session)
+        audit_paritition_nodes, insert_node, node_to_stop, query_to_fail, query_fail_count = self._test_insert_failure_doesnt_report_success_assign_nodes(session=session)
 
         # TODO: remove the loop when scylladb#24473 is fixed
         # We call get_host_id only to cache host_id
@@ -1231,8 +1240,8 @@ class TestCQLAudit(AuditTester):
             # If any audit mode is not done yet, continue polling.
             all_modes_done = True
             for mode, rows in rows_dict.items():
-                rows_with_error = list(filter(lambda r: r.error, rows))
-                if len(rows_with_error) == 6:
+                rows_with_error = [row for row in rows if row.error and row.operation == query_to_fail]
+                if len(rows_with_error) == query_fail_count:
                     logger.info(f"audit mode {mode} log updated after {i} iterations ({i / 10}s)")
                     assert rows_with_error[0].error is True
                     assert rows_with_error[0].consistency == "THREE"

test/cluster/mv/test_mv_building.py

@@ -16,16 +16,26 @@ from test.cluster.util import get_topology_coordinator, new_test_keyspace, recon
 
 logger = logging.getLogger(__name__)
 
-# This test makes sure that view building is done mainly in the streaming scheduling group
-# and not the gossip scheduling group. We do that by measuring the time each group was
-# busy during the view building process and confirming that the gossip group was busy
-# much less than the streaming group.
-# Reproduces https://github.com/scylladb/scylladb/issues/21232
+# This test makes sure that view building is done mainly in the streaming
+# scheduling group. We check that by grepping all relevant logs in TRACE mode
+# and verifying that they come from the streaming scheduling group.
+#
+# For more context, see: https://github.com/scylladb/scylladb/issues/21232.
+# This test reproduces the issue in non-tablet mode.
 @pytest.mark.asyncio
 @skip_mode('debug', 'the test needs to do some work which takes too much time in debug mode')
 async def test_view_building_scheduling_group(manager: ManagerClient):
-    server = await manager.server_add()
+    # Note: The view building coordinator works in the gossiping scheduling group,
+    #       and we intentionally omit it here.
+    # Note: We include "view" for keyspaces that don't use the view building coordinator
+    #       and will follow the legacy path instead.
+    loggers = ["view_building_worker", "view_consumer", "view_update_generator", "view"]
+    # Flatten the list of lists.
+    cmdline = sum([["--logger-log-level", f"{logger}=trace"] for logger in loggers], [])
+
+    server = await manager.server_add(cmdline=cmdline)
     cql = manager.get_cql()
+
     async with new_test_keyspace(manager, "WITH replication = {'class': 'NetworkTopologyStrategy', 'replication_factor': 1}") as ks:
         await cql.run_async(f"CREATE TABLE {ks}.tab (p int, c int, PRIMARY KEY (p, c))")
 
@@ -35,21 +45,30 @@ async def test_view_building_scheduling_group(manager: ManagerClient):
             batch = "BEGIN UNLOGGED BATCH\n" + "\n".join(inserts) + "\nAPPLY BATCH\n"
             await manager.cql.run_async(batch)
 
-        metrics_before = await manager.metrics.query(server.ip_addr)
-        ms_gossip_before = metrics_before.get('scylla_scheduler_runtime_ms', {'group': 'gossip'})
-        ms_streaming_before = metrics_before.get('scylla_scheduler_runtime_ms', {'group': 'streaming'})
+        log = await manager.server_open_log(server.server_id)
+        mark = await log.mark()
 
         await cql.run_async(f"CREATE MATERIALIZED VIEW {ks}.mv AS SELECT p, c FROM {ks}.tab WHERE p IS NOT NULL AND c IS NOT NULL PRIMARY KEY (c, p)")
         await wait_for_view(cql, 'mv', 1)
 
-        metrics_after = await manager.metrics.query(server.ip_addr)
-        ms_gossip_after = metrics_after.get('scylla_scheduler_runtime_ms', {'group': 'gossip'})
-        ms_streaming_after = metrics_after.get('scylla_scheduler_runtime_ms', {'group': 'streaming'})
-        ms_streaming = ms_streaming_after - ms_streaming_before
-        ms_statement = ms_gossip_after - ms_gossip_before
-        ratio = ms_statement / ms_streaming
-        print(f"ms_streaming: {ms_streaming}, ms_statement: {ms_statement}, ratio: {ratio}")
-        assert ratio < 0.1
+        logger_alternative = "|".join(loggers)
+        pattern = rf"\[shard [0-9]+:(.+)\] ({logger_alternative}) - "
+
+        results = await log.grep(pattern, from_mark=mark)
+        # Sanity check. If there are no logs, something's wrong.
+        assert len(results) > 0
+
+        # In case of non-tablet keyspaces, we won't use the view building coordinator.
+        # Instead, view updates will follow the legacy path. Along the way, we'll observe
+        # this message, which will be printed using another scheduling group, so let's
+        # filter it out.
+        predicate = lambda result: f"Building view {ks}.mv, starting at token" not in result[0]
+        results = list(filter(predicate, results))
+
+        # Take the first parenthesized match for each result, i.e. the scheduling group.
+        sched_groups = [matches[1] for _, matches in results]
+
+        assert all(sched_group == "strm" for sched_group in sched_groups)
 
 # A sanity check test ensures that starting and shutting down Scylla when view building is
 # disabled is conducted properly and we don't run into any issues.

test/cluster/test_alternator.py

@@ -25,12 +25,14 @@ import json
 from cassandra.auth import PlainTextAuthProvider
 import threading
 import random
+import re
 
 from test.cluster.util import get_replication
 from test.pylib.manager_client import ManagerClient
 from test.pylib.util import wait_for
 from test.pylib.tablets import get_all_tablet_replicas
 from test.cluster.conftest import skip_mode
+from test.pylib.tablets import get_tablet_replica
 
 logger = logging.getLogger(__name__)
 
@@ -969,3 +971,118 @@ async def test_alternator_concurrent_rmw_same_partition_different_server(manager
             t.join()
     finally:
         table.delete()
+
+
+@pytest.mark.asyncio
+@skip_mode('release', 'error injections are not supported in release mode')
+async def test_alternator_invalid_shard_for_lwt(manager: ManagerClient):
+    """
+    Reproducer for issue #27353.
+
+    LWT requires that storage_proxy::cas() is invoked on a valid shard — the one
+    returned by sharder.try_get_shard_for_reads() for a tablets-based table.
+
+    The bug: if the current shard is invalid and we jump to the valid shard, that
+    new shard may become invalid again by the time we attempt to capture the ERM.
+    This leads to a failure of the CAS path.
+
+    The fix: retry the validity check and jump again if the current shard is already
+    invalid. We should exit the loop once the shard is valid *and* we hold a strong pointer
+    to the ERM — which prevents further tablet movements until the ERM is released.
+
+    This problem is specific to BatchWriteItem; other commands are already handled
+    correctly.
+    """
+    config = alternator_config.copy()
+    config['alternator_write_isolation'] = 'always_use_lwt'
+    cmdline = [
+        '--logger-log-level', 'alternator-executor=trace',
+        '--logger-log-level', 'alternator_controller=trace',
+        '--logger-log-level', 'paxos=trace'
+    ]
+    server = await manager.server_add(config=config, cmdline=cmdline)
+    alternator = get_alternator(server.ip_addr)
+
+    logger.info("Creating alternator test table")
+    table = alternator.create_table(TableName=unique_table_name(),
+                                    Tags=[{'Key': 'system:initial_tablets', 'Value': '1'}],
+                                    BillingMode='PAY_PER_REQUEST',
+                                    KeySchema=[{'AttributeName': 'p', 'KeyType': 'HASH'}],
+                                    AttributeDefinitions=[{'AttributeName': 'p', 'AttributeType': 'N'}])
+    table_name = table.name
+    ks_name = 'alternator_' + table_name
+    last_token = 7 # Any token works since we have only one tablet
+    
+    (src_host_id, src_shard) = await get_tablet_replica(manager, server, ks_name, table_name, last_token)
+    dst_shard = 0 if src_shard == 1 else 1
+
+    logger.info("Inject 'intranode_migration_streaming_wait'")
+    await manager.api.enable_injection(server.ip_addr,
+                                       "intranode_migration_streaming_wait",
+                                       one_shot=False)
+
+    logger.info("Start tablet migration")
+    intranode_migration_task = asyncio.create_task(
+        manager.api.move_tablet(server.ip_addr, ks_name, table_name,
+                                src_host_id, src_shard,
+                                src_host_id, dst_shard, last_token))
+
+    logger.info("Open server logs")
+    log = await manager.server_open_log(server.server_id)
+
+    logger.info("Wait for intranode_migration_streaming_wait")
+    await log.wait_for("intranode_migration_streaming: waiting")
+
+    logger.info("Inject 'alternator_executor_batch_write_wait'")
+    await manager.api.enable_injection(server.ip_addr,
+                                       "alternator_executor_batch_write_wait",
+                                       one_shot=False,
+                                       parameters={
+                                           'table': table_name,
+                                           'keyspace': ks_name,
+                                           'shard': dst_shard
+                                       })
+    m = await log.mark()
+
+    # Start a background thread, which tries to hit the alternator_executor_batch_write_wait
+    # injection on the destination shard.
+    logger.info("Start a batch_write thread")
+    stop_event = threading.Event()
+    def run_batch():
+        alternator = get_alternator(server.ip_addr)
+        table = alternator.Table(table_name)
+        while not stop_event.is_set():
+            with table.batch_writer() as batch:
+                batch.put_item(Item={'p': 1, 'x': 'hellow world'})
+    t = ThreadWrapper(target=run_batch)
+    t.start()
+
+    logger.info("Waiting for 'alternator_executor_batch_write_wait: hit'")
+    await log.wait_for("alternator_executor_batch_write_wait: hit", from_mark=m)
+
+    # We have a batch request with "streaming" cas_shard on the destination shard.
+    # This means we have already made a decision to jump to the src_shard.
+    # Now we're releasing the tablet migration so that it reaches write_both_read_new and
+    # and invaldiates this decision.
+
+    m = await log.mark()
+    await manager.api.message_injection(server.ip_addr, "intranode_migration_streaming_wait")
+
+    # The next barrier must be for the write_both_read_new, we need a guarantee
+    # that the src_shard observed it
+    logger.info("Waiting for the next barrier")
+    await log.wait_for(re.escape(f"[shard {src_shard}: gms] raft_topology - raft_topology_cmd::barrier_and_drain done"),
+                       from_mark=m)
+
+    # Now we have a guarantee that a new barrier succeeded on the src_shard,
+    # this means the src_shard has already transitioned to write_both_read_new,
+    # and our batch write will have to jump back to the destination shard.
+
+    logger.info("Release the 'alternator_executor_batch_write_wait'")
+    await manager.api.message_injection(server.ip_addr, "alternator_executor_batch_write_wait")
+
+    logger.info("Waiting for migratino task to finish")
+    await intranode_migration_task
+
+    stop_event.set()
+    t.join()

test/cluster/test_incremental_repair.py

@@ -220,14 +220,14 @@ async def test_tablet_repair_sstable_skipped_read_metrics(manager: ManagerClient
 
     await insert_keys(cql, ks, 0, 100)
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     skipped_bytes = get_incremental_repair_sst_skipped_bytes(servers[0])
     read_bytes = get_incremental_repair_sst_read_bytes(servers[0])
     # Nothing to skip. Repair all data.
     assert skipped_bytes == 0
     assert read_bytes > 0
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     skipped_bytes2 = get_incremental_repair_sst_skipped_bytes(servers[0])
     read_bytes2 = get_incremental_repair_sst_read_bytes(servers[0])
     # Skip all. Nothing to repair
@@ -236,7 +236,7 @@ async def test_tablet_repair_sstable_skipped_read_metrics(manager: ManagerClient
 
     await insert_keys(cql, ks, 200, 300)
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     skipped_bytes3 = get_incremental_repair_sst_skipped_bytes(servers[0])
     read_bytes3 = get_incremental_repair_sst_read_bytes(servers[0])
     # Both skipped and read bytes should grow
@@ -272,7 +272,7 @@ async def test_tablet_incremental_repair(manager: ManagerClient):
     assert get_sstables_repaired_at(map0, token) == sstables_repaired_at
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     map1 = await load_tablet_sstables_repaired_at(manager, cql, servers[0], hosts[0], table_id)
     logging.info(f'map1={map1}')
     # Check sstables_repaired_at is increased by 1
@@ -288,7 +288,7 @@ async def test_tablet_incremental_repair(manager: ManagerClient):
         assert len(enable) == 1
 
     # Second repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     map2 = await load_tablet_sstables_repaired_at(manager, cql, servers[0], hosts[0], table_id)
     logging.info(f'map2={map2}')
     # Check sstables_repaired_at is increased by 1
@@ -313,7 +313,7 @@ async def test_tablet_incremental_repair_error(manager: ManagerClient):
     # Repair should not finish with error
     await inject_error_on(manager, "repair_tablet_fail_on_rpc_call", servers)
     try:
-        await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, timeout=10)
+        await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental', timeout=10)
         assert False # Check the tablet repair is not supposed to finish
     except TimeoutError:
         logger.info("Repair timeout as expected")
@@ -329,7 +329,7 @@ async def do_tablet_incremental_repair_and_ops(manager: ManagerClient, ops: str)
     servers, cql, hosts, ks, table_id, logs, repaired_keys, unrepaired_keys, current_key, token = await preapre_cluster_for_incremental_repair(manager, nr_keys)
     token = -1
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     # 1 add 0 skip 1 mark
     for log in logs:
         sst_add, sst_skip, sst_mark = await get_sst_status("First", log)
@@ -355,7 +355,7 @@ async def do_tablet_incremental_repair_and_ops(manager: ManagerClient, ops: str)
         else:
             assert False # Wrong ops
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
 
     # 1 add 1 skip 1 mark
     for log in logs:
@@ -394,7 +394,7 @@ async def test_tablet_incremental_repair_and_minor(manager: ManagerClient):
         await manager.api.disable_autocompaction(server.ip_addr, ks, 'test')
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
 
     # Insert more keys
     await insert_keys(cql, ks, current_key, current_key + nr_keys)
@@ -402,7 +402,7 @@ async def test_tablet_incremental_repair_and_minor(manager: ManagerClient):
     current_key += nr_keys
 
     # Second repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
 
     # Insert more keys and flush to get 2 more sstables
     for _ in range(2):
@@ -436,7 +436,7 @@ async def do_test_tablet_incremental_repair_with_split_and_merge(manager, do_spl
     servers, cql, hosts, ks, table_id, logs, repaired_keys, unrepaired_keys, current_key, token = await preapre_cluster_for_incremental_repair(manager, nr_keys)
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 1
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 1
     await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 2
 
     # Insert more keys
@@ -445,7 +445,7 @@ async def do_test_tablet_incremental_repair_with_split_and_merge(manager, do_spl
     current_key += nr_keys
 
     # Second repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 3
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 3
 
     # Insert more keys and flush to get 2 more sstables
     for _ in range(2):
@@ -505,7 +505,7 @@ async def test_tablet_incremental_repair_existing_and_repair_produced_sstable(ma
 
     await manager.server_start(servers[1].server_id)
 
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
 
     scylla_path = get_scylla_path(cql)
 
@@ -521,8 +521,8 @@ async def test_tablet_incremental_repair_merge_higher_repaired_at_number(manager
     servers, cql, hosts, ks, table_id, logs, repaired_keys, unrepaired_keys, current_key, token = await preapre_cluster_for_incremental_repair(manager, nr_keys)
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 1
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 2
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 1
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 2
 
     # Insert more keys
     await insert_keys(cql, ks, current_key, current_key + nr_keys)
@@ -532,7 +532,7 @@ async def test_tablet_incremental_repair_merge_higher_repaired_at_number(manager
     # Second repair
     await inject_error_on(manager, "repair_tablet_no_update_sstables_repair_at", servers)
     # some sstable repaired_at = 3, but sstables_repaired_at = 2
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 2
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 2
     await inject_error_off(manager, "repair_tablet_no_update_sstables_repair_at", servers)
 
     scylla_path = get_scylla_path(cql)
@@ -561,8 +561,8 @@ async def test_tablet_incremental_repair_merge_correct_repaired_at_number_after_
     servers, cql, hosts, ks, table_id, logs, repaired_keys, unrepaired_keys, current_key, token = await preapre_cluster_for_incremental_repair(manager, nr_keys)
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 1
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 2
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 1
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 2
 
     # Insert more keys
     await insert_keys(cql, ks, current_key, current_key + nr_keys)
@@ -574,7 +574,7 @@ async def test_tablet_incremental_repair_merge_correct_repaired_at_number_after_
     last_tokens = [t.last_token for t in replicas]
     for t in last_tokens[0::2]:
         logging.info(f"Start repair for token={t}");
-        await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", t) # sstables_repaired_at 3
+        await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", t, incremental_mode='incremental') # sstables_repaired_at 3
 
     scylla_path = get_scylla_path(cql)
 
@@ -595,7 +595,7 @@ async def do_test_tablet_incremental_repair_merge_error(manager, error):
     servers, cql, hosts, ks, table_id, logs, repaired_keys, unrepaired_keys, current_key, token = await preapre_cluster_for_incremental_repair(manager, nr_keys, cmdline)
 
     # First repair
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token) # sstables_repaired_at 1
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental') # sstables_repaired_at 1
 
     # Insert more keys
     await insert_keys(cql, ks, current_key, current_key + nr_keys)
@@ -659,13 +659,18 @@ async def test_tablet_repair_with_incremental_option(manager: ManagerClient):
         assert read1 == 0
         assert skip2 == 0
         assert read2 > 0
-    await do_repair_and_check(None, 1, rf'Starting tablet repair by API .* incremental_mode=incremental.*', check1)
+    await do_repair_and_check('incremental', 1, rf'Starting tablet repair by API .* incremental_mode=incremental.*', check1)
 
     def check2(skip1, read1, skip2, read2):
         assert skip1 == skip2
         assert read1 == read2
     await do_repair_and_check('disabled', 0, rf'Starting tablet repair by API .* incremental_mode=disabled.*', check2)
 
+    # FIXME: Incremental repair is disabled by default due to
+    # https://github.com/scylladb/scylladb/issues/26041 and
+    # https://github.com/scylladb/scylladb/issues/27414
+    await do_repair_and_check(None, 0, rf'Starting tablet repair by API .* incremental_mode=disabled.*', check2)
+
     def check3(skip1, read1, skip2, read2):
         assert skip1 < skip2
         assert read1 == read2
@@ -677,14 +682,14 @@ async def test_tablet_repair_with_incremental_option(manager: ManagerClient):
     await do_repair_and_check('full', 1, rf'Starting tablet repair by API .* incremental_mode=full.*', check4)
 
 @pytest.mark.asyncio
-async def test_tablet_repair_tablet_time_metrics(manager: ManagerClient):
+async def test_incremental_repair_tablet_time_metrics(manager: ManagerClient):
     servers, _, _, ks, _, _, _, _, _, token = await preapre_cluster_for_incremental_repair(manager)
     time1 = 0
     time2 = 0
 
     for s in servers:
         time1 += get_repair_tablet_time_ms(s)
-    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token)
+    await manager.api.tablet_repair(servers[0].ip_addr, ks, "test", token, incremental_mode='incremental')
     for s in servers:
         time2 += get_repair_tablet_time_ms(s)
 
@@ -694,7 +699,7 @@ async def test_tablet_repair_tablet_time_metrics(manager: ManagerClient):
 # Reproducer for https://github.com/scylladb/scylladb/issues/26346
 @pytest.mark.asyncio
 @skip_mode('release', 'error injections are not supported in release mode')
-async def test_repair_finishes_when_tablet_skips_end_repair_stage(manager):
+async def test_incremental_repair_finishes_when_tablet_skips_end_repair_stage(manager):
     servers = await manager.servers_add(3, auto_rack_dc="dc1")
 
     async with new_test_keyspace(manager, "WITH replication = {'class': 'NetworkTopologyStrategy', 'replication_factor': 3} AND tablets = {'initial': 1}") as ks:
@@ -719,7 +724,7 @@ async def test_repair_finishes_when_tablet_skips_end_repair_stage(manager):
 
 @pytest.mark.asyncio
 @skip_mode('release', 'error injections are not supported in release mode')
-async def test_repair_rejoin_do_tablet_operation(manager):
+async def test_incremental_repair_rejoin_do_tablet_operation(manager):
     cmdline = ['--logger-log-level', 'raft_topology=debug']
     servers = await manager.servers_add(3, auto_rack_dc="dc1", cmdline=cmdline)
 

test/cluster/test_ip_mappings.py

@@ -10,6 +10,7 @@ import logging
 
 from test.pylib.rest_client import inject_error_one_shot
 from test.cluster.util import new_test_keyspace
+from test.pylib.util import gather_safely
 
 logger = logging.getLogger(__name__)
 
@@ -33,25 +34,12 @@ async def test_broken_bootstrap(manager: ManagerClient):
         except Exception:
             pass
 
-        await manager.server_stop(server_b.server_id)
-        await manager.server_stop(server_a.server_id)
-
-        stop_event = asyncio.Event()
-        async def worker():
-            logger.info("Worker started")
-            while not stop_event.is_set():
-                for i in range(100):
-                    await manager.cql.run_async(f"INSERT INTO {table} (a, b) VALUES ({i}, {i})")
-                    response = await manager.cql.run_async(f"SELECT * FROM {table} WHERE a = {i}")
-                    assert response[0].b == i
-                await asyncio.sleep(0.1)
-            logger.info("Worker stopped")
+        await gather_safely(*(manager.server_stop(srv.server_id) for srv in [server_a, server_b]))
 
         await manager.server_start(server_a.server_id)
         await manager.driver_connect()
 
-        worker_task = asyncio.create_task(worker())
-
-        await asyncio.sleep(20)
-        stop_event.set()
-        await worker_task
+        for i in range(100):
+            await manager.cql.run_async(f"INSERT INTO {table} (a, b) VALUES ({i}, {i})")
+            response = await manager.cql.run_async(f"SELECT * FROM {table} WHERE a = {i}")
+            assert response[0].b == i

test/cluster/test_tablets2.py

@@ -4,7 +4,8 @@
 # SPDX-License-Identifier: LicenseRef-ScyllaDB-Source-Available-1.0
 #
 from typing import Any
-from cassandra.query import ConsistencyLevel
+from cassandra.query import ConsistencyLevel, SimpleStatement
+from cassandra.policies import FallthroughRetryPolicy
 
 from test.pylib.internal_types import HostID, ServerInfo, ServerNum
 from test.pylib.manager_client import ManagerClient
@@ -1596,7 +1597,7 @@ async def test_truncate_during_topology_change(manager: ManagerClient):
         async def truncate_table():
             await asyncio.sleep(10)
             logger.info("Executing truncate during bootstrap")
-            await cql.run_async(f"TRUNCATE {ks}.test USING TIMEOUT 1m")
+            await cql.run_async(SimpleStatement(f"TRUNCATE {ks}.test USING TIMEOUT 4m", retry_policy=FallthroughRetryPolicy()))
 
         truncate_task = asyncio.create_task(truncate_table())
         logger.info("Adding fourth node")

test/cqlpy/nodetool.py

@@ -67,11 +67,11 @@ nodetool_cmd.conf = False
 
 # Run the external "nodetool" executable (can be overridden by the NODETOOL
 # environment variable). Only call this if the REST API doesn't work.
-def run_nodetool(cql, *args):
+def run_nodetool(cql, *args, **subprocess_kwargs):
     # TODO: We may need to change this function or its callers to add proper
     # support for testing on multi-node clusters.
     host = cql.cluster.contact_points[0]
-    subprocess.run([nodetool_cmd(), '-h', host, *args])
+    return subprocess.run([nodetool_cmd(), '-h', host, *args], **subprocess_kwargs)
 
 def flush(cql, table):
     ks, cf = table.split('.')
@@ -157,6 +157,28 @@ def disablebinary(cql):
     else:
         run_nodetool(cql, "disablebinary")
 
+def getlogginglevel(cql, logger):
+    if has_rest_api(cql):
+        resp = requests.get(f'{rest_api_url(cql)}/system/logger/{logger}')
+        if resp.ok:
+            return resp.text.strip()
+        raise RuntimeError(f"failed to fetch logging level for {logger}: {resp.status_code} {resp.text}")
+
+    result = run_nodetool(
+        cql,
+        "getlogginglevels",
+        capture_output=True,
+        text=True,
+        check=True,
+    )
+    for line in result.stdout.splitlines():
+        stripped = line.strip()
+        parts = stripped.split()
+        if len(parts) >= 2 and parts[0] == logger:
+            return parts[-1]
+
+    raise RuntimeError(f"logger {logger} not found in getlogginglevels output")
+
 def setlogginglevel(cql, logger, level):
     if has_rest_api(cql):
         requests.post(f'{rest_api_url(cql)}/system/logger/{logger}', params={'level': level})

test/cqlpy/test_protocol_exceptions.py

@@ -10,6 +10,7 @@ import re
 import requests
 import socket
 import struct
+from test.cqlpy import nodetool
 from test.cqlpy.util import cql_session
 
 def get_protocol_error_metrics(host) -> int:
@@ -58,11 +59,50 @@ def try_connect(host, port, creds, protocol_version):
     with cql_with_protocol(host, port, creds, protocol_version) as session:
         return 1 if session else 0
 
+@pytest.fixture
+def debug_exceptions_logging(request, cql):
+    def _read_level() -> str | None:
+        try:
+            level = nodetool.getlogginglevel(cql, "exception")
+            if level:
+                level = level.strip().strip('"').lower()
+            return level
+        except Exception as exc:
+            print(f"Failed to read exception logger level: {exc}")
+            return None
+
+    def _set_and_verify(level: str) -> bool:
+        try:
+            nodetool.setlogginglevel(cql, "exception", level)
+        except Exception as exc:
+            print(f"Failed to set exception logger level to '{level}': {exc}")
+            return False
+
+        observed = _read_level()
+        if observed == level:
+            return True
+
+        print(f"Exception logger level observed as '{observed}' while expecting '{level}'")
+        return False
+
+    def _restore_logging():
+        if not enabled and previous_level is None:
+            return
+
+        target_level = previous_level or "info"
+        _set_and_verify(target_level)
+
+    previous_level = _read_level()
+    enabled = _set_and_verify("debug")
+
+    yield
+    _restore_logging()
+
 # If there is a protocol version mismatch, the server should
 # raise a protocol error, which is counted in the metrics.
-def test_protocol_version_mismatch(scylla_only, request, host):
-    run_count = 100
-    cpp_exception_threshold = 10
+def test_protocol_version_mismatch(scylla_only, debug_exceptions_logging, request, host):
+    run_count = 200
+    cpp_exception_threshold = 20
 
     cpp_exception_metrics_before = get_cpp_exceptions_metrics(host)
     protocol_exception_metrics_before = get_protocol_error_metrics(host)
@@ -244,8 +284,8 @@ def _protocol_error_impl(
         s.close()
 
 def _test_impl(host, flag):
-    run_count = 100
-    cpp_exception_threshold = 10
+    run_count = 200
+    cpp_exception_threshold = 20
 
     cpp_exception_metrics_before = get_cpp_exceptions_metrics(host)
     protocol_exception_metrics_before = get_protocol_error_metrics(host)
@@ -267,47 +307,47 @@ def no_ssl(request):
     yield
 
 # Malformed BATCH with an invalid kind triggers a protocol error.
-def test_invalid_kind_in_batch_message(scylla_only, no_ssl, host):
+def test_invalid_kind_in_batch_message(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_bad_batch")
 
 # Send OPTIONS during AUTHENTICATE to trigger auth-state error.
-def test_unexpected_message_during_auth(scylla_only, no_ssl, host):
+def test_unexpected_message_during_auth(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_unexpected_auth")
 
 # STARTUP with an invalid/missing string-map entry should produce a protocol error.
-def test_process_startup_invalid_string_map(scylla_only, no_ssl, host):
+def test_process_startup_invalid_string_map(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_startup_invalid_string_map")
 
 # STARTUP with unknown COMPRESSION option should produce a protocol error.
-def test_unknown_compression_algorithm(scylla_only, no_ssl, host):
+def test_unknown_compression_algorithm(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_unknown_compression")
 
 # QUERY long-string truncation: declared length > provided bytes triggers protocol error.
-def test_process_query_internal_malformed_query(scylla_only, no_ssl, host):
+def test_process_query_internal_malformed_query(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_query_internal_malformed_query")
 
 # QUERY options malformed: PAGE_SIZE flag set but page_size truncated triggers protocol error.
-def test_process_query_internal_fail_read_options(scylla_only, no_ssl, host):
+def test_process_query_internal_fail_read_options(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_query_internal_fail_read_options")
 
 # PREPARE long-string truncation: declared length > provided bytes triggers protocol error.
-def test_process_prepare_malformed_query(scylla_only, no_ssl, host):
+def test_process_prepare_malformed_query(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_prepare_malformed_query")
 
 # EXECUTE cache-key malformed: short-bytes length > provided bytes triggers protocol error.
-def test_process_execute_internal_malformed_cache_key(scylla_only, no_ssl, host):
+def test_process_execute_internal_malformed_cache_key(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_execute_internal_malformed_cache_key")
 
 # REGISTER malformed string list: declared string length > provided bytes triggers protocol error.
-def test_process_register_malformed_string_list(scylla_only, no_ssl, host):
+def test_process_register_malformed_string_list(scylla_only, no_ssl, debug_exceptions_logging, host):
     _test_impl(host, "trigger_process_register_malformed_string_list")
 
 # Test if the protocol exceptions do not decrease after running the test happy path.
 # This is to ensure that the protocol exceptions are not cleared or reset
 # during the test execution.
-def test_no_protocol_exceptions(scylla_only, no_ssl, host):
-    run_count = 100
-    cpp_exception_threshold = 10
+def test_no_protocol_exceptions(scylla_only, no_ssl, debug_exceptions_logging, host):
+    run_count = 200
+    cpp_exception_threshold = 20
 
     cpp_exception_metrics_before = get_cpp_exceptions_metrics(host)
     protocol_exception_metrics_before = get_protocol_error_metrics(host)

test/lib/sstable_utils.hh

@@ -163,6 +163,11 @@ public:
         _sst->_shards.push_back(this_shard_id());
     }
 
+    void set_first_and_last_keys(const dht::decorated_key& first_key, const dht::decorated_key& last_key) {
+        _sst->_first = first_key;
+        _sst->_last = last_key;
+    }
+
     void rewrite_toc_without_component(component_type component) {
         SCYLLA_ASSERT(component != component_type::TOC);
         _sst->_recognized_components.erase(component);

test/pylib/resource_gather.py

@@ -109,6 +109,7 @@ class ResourceGather(ABC):
         except subprocess.TimeoutExpired:
             logger.critical(f"Process {args} timed out")
             p.kill()
+            p.communicate()
         except KeyboardInterrupt:
             p.kill()
             raise

test/pylib/scylla_cluster.py

@@ -789,7 +789,7 @@ class ScyllaServer:
 
         while time.time() < self.start_time + self.TOPOLOGY_TIMEOUT and not self.stop_event.is_set():
             assert self.cmd is not None
-            if self.cmd.returncode:
+            if self.cmd.returncode is not None:
                 self.cmd = None
                 if expected_error is not None:
                     with self.log_filename.open("r", encoding="utf-8") as log_file:

tools/scylla-nodetool.cc

@@ -654,7 +654,7 @@ void cluster_repair_operation(scylla_rest_client& client, const bpo::variables_m
             for (const auto& table : tables.empty() ? ks_to_cfs[keyspace] : tables) {
                 repair_params["table"] = table;
                 try {
-                    sstring task_id = client.post("/storage_service/tablets/repair", repair_params).GetObject()["tablet_task_id"].GetString();
+                    sstring task_id = rjson::to_sstring(client.post("/storage_service/tablets/repair", repair_params).GetObject()["tablet_task_id"]);
 
                     log("Starting repair with task_id={} keyspace={} table={}", task_id, keyspace, table);
 

tools/toolchain/Dockerfile

@@ -1,4 +1,4 @@
-FROM docker.io/fedora:42
+FROM registry.fedoraproject.org/fedora:43
 
 ARG CLANG_BUILD="SKIP"
 ARG CLANG_ARCHIVES

tools/toolchain/image

@@ -1 +1 @@
-docker.io/scylladb/scylla-toolchain:fedora-42-20251122
+docker.io/scylladb/scylla-toolchain:fedora-43-20251208

tools/toolchain/optimized_clang.sh

@@ -65,7 +65,7 @@ SCYLLA_BUILD_DIR_FULLPATH="${SCYLLA_DIR}"/"${SCYLLA_BUILD_DIR}"
 SCYLLA_NINJA_FILE_FULLPATH="${SCYLLA_DIR}"/"${SCYLLA_NINJA_FILE}"
 
 # Which LLVM release to build in order to compile Scylla
-LLVM_CLANG_TAG=20.1.8
+LLVM_CLANG_TAG=21.1.6
 
 CLANG_ARCHIVE=$(cd "${SCYLLA_DIR}" && realpath -m "${CLANG_ARCHIVE}")
 
@@ -123,7 +123,7 @@ _get_distribution_components() {
                 continue
                 ;;
         esac
-        echo $target
+        echo "$target"
     done
 }
 
@@ -164,7 +164,7 @@ if [[ "${CLANG_BUILD}" = "INSTALL" ]]; then
 
     echo "[clang-stage3] build the compiler applied CSPGO profile"
     cd "${CLANG_BUILD_DIR}"
-    llvm-profdata merge build/csprofiles/default_*.profraw -output=csir.prof
+    llvm-profdata merge build/profiles/csir-*.profraw -output=csir.prof
     llvm-profdata merge ir.prof csir.prof -output=combined.prof
     rm -rf build
     # linker flags are needed for BOLT
@@ -186,7 +186,3 @@ if [[ $? -ne 0 ]]; then
 fi
 set -e
 tar -C / -xpzf "${CLANG_ARCHIVE}"
-dnf remove -y clang clang-libs
-# above package removal might have removed those symbolic links, which will cause ccache not to work later on. Manually restore them.
-ln -sf /usr/bin/ccache /usr/lib64/ccache/clang
-ln -sf /usr/bin/ccache /usr/lib64/ccache/clang++

transport/generic_server.cc

@@ -29,11 +29,8 @@ class counted_data_source_impl : public data_source_impl {
         if (_cpu_concurrency.stopped) {
             return fun();
         }
-        return futurize_invoke([this] () {
-            _cpu_concurrency.units.return_all();
-        }).then([fun = std::move(fun)] () {
-            return fun();
-        }).finally([this] () {
+        _cpu_concurrency.units.return_all();
+        return fun().finally([this] () {
             _cpu_concurrency.units.adopt(consume_units(_cpu_concurrency.semaphore, 1));
         });
     };
@@ -60,11 +57,8 @@ class counted_data_sink_impl : public data_sink_impl {
         if (_cpu_concurrency.stopped) {
             return fun();
         }
-        return futurize_invoke([this] () {
-            _cpu_concurrency.units.return_all();
-        }).then([fun = std::move(fun)] () mutable {
-            return fun();
-        }).finally([this] () {
+        _cpu_concurrency.units.return_all();
+        return fun().finally([this] () {
             _cpu_concurrency.units.adopt(consume_units(_cpu_concurrency.semaphore, 1));
         });
     };

types/types.hh

@@ -832,6 +832,12 @@ to_bytes(bytes_view x) {
     return bytes(x.begin(), x.size());
 }
 
+inline
+bytes
+to_bytes(std::string_view x) {
+    return to_bytes(to_bytes_view(x));
+}
+
 inline
 bytes_opt
 to_bytes_opt(bytes_view_opt bv) {

utils/error_injection.hh

@@ -204,7 +204,7 @@ public:
 
     public:
         template <typename Clock, typename Duration>
-        future<> wait_for_message(std::chrono::time_point<Clock, Duration> timeout, abort_source* as = nullptr) {
+        future<> wait_for_message(std::chrono::time_point<Clock, Duration> timeout, abort_source* as = nullptr, std::source_location loc = std::source_location::current()) {
             if (!_shared_data) {
                 on_internal_error(errinj_logger, "injection_shared_data is not initialized");
             }
@@ -234,7 +234,8 @@ public:
                 throw;
             }
             catch (const std::exception& e) {
-                on_internal_error(errinj_logger, "Error injection wait_for_message timeout: " + std::string(e.what()));
+                on_internal_error(errinj_logger, fmt::format("Error injection [{}] wait_for_message timeout: Called from `{}` @ {}:{}:{:d}: {}",
+                        _shared_data->injection_name, loc.function_name(), loc.file_name(), loc.line(), loc.column(), e.what()));
             }
             ++_read_messages_counter;
         }

utils/rjson.hh

@@ -174,13 +174,6 @@ future<> print_with_extra_array(const rjson::value& value,
     seastar::output_stream<char>& os,
     size_t max_nested_level = default_max_nested_level);
 
-// Returns a string_view to the string held in a JSON value (which is
-// assumed to hold a string, i.e., v.IsString() == true). This is a view
-// to the existing data - no copying is done.
-inline std::string_view to_string_view(const rjson::value& v) {
-    return std::string_view(v.GetString(), v.GetStringLength());
-}
-
 // Copies given JSON value - involves allocation
 rjson::value copy(const rjson::value& value);
 
@@ -236,6 +229,27 @@ rjson::value parse_yieldable(chunked_content&&, size_t max_nested_level = defaul
 rjson::value from_string(const char* str, size_t size);
 rjson::value from_string(std::string_view view);
 
+// Returns a string_view to the string held in a JSON value (which is
+// assumed to hold a string, i.e., v.IsString() == true). This is a view
+// to the existing data - no copying is done.
+inline std::string_view to_string_view(const rjson::value& v) {
+    return std::string_view(v.GetString(), v.GetStringLength());
+}
+
+// Those functions must be called on json string object.
+// They make a copy of underlying data so it's safe to destroy
+// rjson::value afterwards.
+//
+// Rapidjson's GetString method alone is not good enough
+// for string conversion because it needs to scan the string
+// unnecessarily and GetStringLength could be used to avoid that.
+inline sstring to_sstring(const rjson::value& str) {
+    return sstring(str.GetString(), str.GetStringLength());
+}
+inline std::string to_string(const rjson::value& str) {
+    return std::string(str.GetString(), str.GetStringLength());
+}
+
 // Returns a pointer to JSON member if it exists, nullptr otherwise
 rjson::value* find(rjson::value& value, std::string_view name);
 const rjson::value* find(const rjson::value& value, std::string_view name);
@@ -377,7 +391,7 @@ rjson::value from_string_map(const std::map<sstring, sstring>& map);
 sstring quote_json_string(const sstring& value);
 
 inline bytes base64_decode(const value& v) {
-    return ::base64_decode(std::string_view(v.GetString(), v.GetStringLength()));
+    return ::base64_decode(to_string_view(v));
 }
 
 // A writer which allows writing json into an std::ostream in a streaming manner.

utils/s3/credentials_providers/instance_profile_credentials_provider.cc

@@ -10,6 +10,7 @@
 #include "utils/http.hh"
 #include "utils/s3/client.hh"
 #include "utils/s3/default_aws_retry_strategy.hh"
+#include "utils/rjson.hh"
 #include <rapidjson/document.h>
 #include <rapidjson/error/en.h>
 #include <seastar/http/client.hh>
@@ -72,7 +73,7 @@ future<> instance_profile_credentials_provider::update_credentials() {
 }
 
 s3::aws_credentials instance_profile_credentials_provider::parse_creds(const sstring& creds_response) {
-    rapidjson::Document document;
+    rjson::document document;
     document.Parse(creds_response.data());
 
     if (document.HasParseError()) {
@@ -81,9 +82,9 @@ s3::aws_credentials instance_profile_credentials_provider::parse_creds(const sst
     }
 
     // Retrieve credentials
-    return {.access_key_id = document["AccessKeyId"].GetString(),
-            .secret_access_key = document["SecretAccessKey"].GetString(),
-            .session_token = document["Token"].GetString(),
+    return {.access_key_id = rjson::to_string(document["AccessKeyId"]),
+            .secret_access_key = rjson::to_string(document["SecretAccessKey"]),
+            .session_token = rjson::to_string(document["Token"]),
             // Set the expiration to one minute earlier to ensure credentials are renewed slightly before they expire
             .expires_at = seastar::lowres_clock::now() + std::chrono::seconds(session_duration - 60)};
 }

vector_search/client.cc

@@ -152,7 +152,7 @@ seastar::future<bool> client::check_status() {
     }
     auto resp = co_await std::move(f);
     auto json = rjson::parse(std::move(resp.content));
-    co_return json.IsString() && json.GetString() == std::string_view("SERVING");
+    co_return json.IsString() && rjson::to_string_view(json) == "SERVING";
 }
 
 seastar::future<> client::close() {