d8826acaa3
The code was incorrectly passing a data_value of type bytes due to
implicit conversion of the result of serialize() (bytes_opt) to a
data_value object of type bytes_type via:
data_value(std::optional<NativeType>);
mutation::set_static_cell() accepts a data_value object, which is then
serialized using column's type in abstract_type::decompose(data_value&):
bytes b(bytes::initialized_later(), serialized_size(*this, value._value));
auto i = b.begin();
value.serialize(i);
Notice that serialized_size() is taken from the column type, but
serialization is done using data_value's type. The two types may have
a compatible CQL binary representation, but may differ in native
types. serialized_size() may incorrectly interpret the native type and
come up with the wrong size. If the size is too smaller, we end up with
stack or heap corruption later after serialize().
For example, if the column type is utf8 but value holds bytes, the
size will be wrong because even though both use the basic_sstring
type, they have a different layout due to max_size (15 vs 31).
Fixes #13717
Closes #13787
214 lines
8.7 KiB
C++
214 lines
8.7 KiB
C++
/*
|
|
* Copyright (C) 2023-present ScyllaDB
|
|
*/
|
|
|
|
/*
|
|
* SPDX-License-Identifier: AGPL-3.0-or-later
|
|
*/
|
|
|
|
#include "types/types.hh"
|
|
#include "types/tuple.hh"
|
|
#include "types/set.hh"
|
|
#include "db/system_keyspace.hh"
|
|
#include "schema/schema_builder.hh"
|
|
#include "cql3/query_processor.hh"
|
|
#include "cql3/untyped_result_set.hh"
|
|
#include "replica/database.hh"
|
|
#include "replica/tablets.hh"
|
|
|
|
namespace replica {
|
|
|
|
using namespace locator;
|
|
|
|
static thread_local auto replica_type = tuple_type_impl::get_instance({uuid_type, int32_type});
|
|
static thread_local auto replica_set_type = set_type_impl::get_instance(replica_type, false);
|
|
|
|
schema_ptr make_tablets_schema() {
|
|
// FIXME: Allow UDTs in system keyspace:
|
|
// CREATE TYPE tablet_replica (replica_id uuid, shard int);
|
|
// replica_set_type = frozen<set<tablet_replica>>
|
|
auto id = generate_legacy_id(db::system_keyspace::NAME, db::system_keyspace::TABLETS);
|
|
return schema_builder(db::system_keyspace::NAME, db::system_keyspace::TABLETS, id)
|
|
.with_column("keyspace_name", utf8_type, column_kind::partition_key)
|
|
.with_column("table_id", uuid_type, column_kind::partition_key)
|
|
.with_column("tablet_count", int32_type, column_kind::static_column)
|
|
.with_column("table_name", utf8_type, column_kind::static_column)
|
|
.with_column("last_token", long_type, column_kind::clustering_key)
|
|
.with_column("replicas", replica_set_type)
|
|
.with_column("new_replicas", replica_set_type)
|
|
.with_column("stage", utf8_type)
|
|
.with_version(db::system_keyspace::generate_schema_version(id))
|
|
.build();
|
|
}
|
|
|
|
future<mutation>
|
|
tablet_map_to_mutation(const tablet_map& tablets, table_id id, const sstring& keyspace_name, const sstring& table_name,
|
|
api::timestamp_type ts) {
|
|
auto s = db::system_keyspace::tablets();
|
|
|
|
auto replicas_to_data_value = [&] (const tablet_replica_set& replicas) {
|
|
std::vector<data_value> result;
|
|
result.reserve(replicas.size());
|
|
for (auto&& replica : replicas) {
|
|
result.emplace_back(make_tuple_value(replica_type, {
|
|
data_value(utils::UUID(replica.host.uuid())),
|
|
data_value(int(replica.shard))
|
|
}));
|
|
}
|
|
return result;
|
|
};
|
|
|
|
auto gc_now = gc_clock::now();
|
|
auto tombstone_ts = ts - 1;
|
|
|
|
mutation m(s, partition_key::from_exploded(*s, {
|
|
data_value(keyspace_name).serialize_nonnull(),
|
|
data_value(id.uuid()).serialize_nonnull()
|
|
}));
|
|
m.partition().apply(tombstone(tombstone_ts, gc_now));
|
|
m.set_static_cell("tablet_count", data_value(int(tablets.tablet_count())), ts);
|
|
m.set_static_cell("table_name", data_value(table_name), ts);
|
|
|
|
tablet_id tid = tablets.first_tablet();
|
|
for (auto&& tablet : tablets.tablets()) {
|
|
auto last_token = tablets.get_last_token(tid);
|
|
auto ck = clustering_key::from_single_value(*s, data_value(dht::token::to_int64(last_token)).serialize_nonnull());
|
|
m.set_clustered_cell(ck, "replicas", make_set_value(replica_set_type, replicas_to_data_value(tablet.replicas)), ts);
|
|
if (auto tr_info = tablets.get_tablet_transition_info(tid)) {
|
|
m.set_clustered_cell(ck, "new_replicas", make_set_value(replica_set_type, replicas_to_data_value(tr_info->next)), ts);
|
|
}
|
|
tid = *tablets.next_tablet(tid);
|
|
co_await coroutine::maybe_yield();
|
|
}
|
|
co_return std::move(m);
|
|
}
|
|
|
|
mutation make_drop_tablet_map_mutation(const sstring& keyspace_name, table_id id, api::timestamp_type ts) {
|
|
auto s = db::system_keyspace::tablets();
|
|
mutation m(s, partition_key::from_exploded(*s, {
|
|
data_value(keyspace_name).serialize_nonnull(),
|
|
data_value(id.uuid()).serialize_nonnull()
|
|
}));
|
|
m.partition().apply(tombstone(ts, gc_clock::now()));
|
|
return m;
|
|
}
|
|
|
|
static
|
|
tablet_replica_set deserialize_replica_set(cql3::untyped_result_set_row::view_type raw_value) {
|
|
tablet_replica_set result;
|
|
auto v = value_cast<set_type_impl::native_type>(
|
|
replica_set_type->deserialize_value(raw_value));
|
|
result.reserve(v.size());
|
|
for (const data_value& replica_v : v) {
|
|
std::vector<data_value> replica_dv = value_cast<tuple_type_impl::native_type>(replica_v);
|
|
result.emplace_back(tablet_replica {
|
|
host_id(value_cast<utils::UUID>(replica_dv[0])),
|
|
shard_id(value_cast<int>(replica_dv[1]))
|
|
});
|
|
}
|
|
return result;
|
|
}
|
|
|
|
future<> save_tablet_metadata(replica::database& db, const tablet_metadata& tm, api::timestamp_type ts) {
|
|
tablet_logger.trace("Saving tablet metadata: {}", tm);
|
|
std::vector<mutation> muts;
|
|
muts.reserve(tm.all_tables().size());
|
|
for (auto&& [id, tablets] : tm.all_tables()) {
|
|
// FIXME: Should we ignore missing tables? Currently doesn't matter because this is only used in tests.
|
|
auto s = db.find_schema(id);
|
|
muts.emplace_back(
|
|
co_await tablet_map_to_mutation(tablets, id, s->ks_name(), s->cf_name(), ts));
|
|
}
|
|
co_await db.apply(freeze(muts), db::no_timeout);
|
|
}
|
|
|
|
future<tablet_metadata> read_tablet_metadata(cql3::query_processor& qp) {
|
|
tablet_metadata tm;
|
|
struct active_tablet_map {
|
|
table_id table;
|
|
tablet_map map;
|
|
tablet_id tid;
|
|
};
|
|
std::optional<active_tablet_map> current;
|
|
|
|
auto process_row = [&] (const cql3::untyped_result_set_row& row) {
|
|
auto table = table_id(row.get_as<utils::UUID>("table_id"));
|
|
|
|
if (!current || current->table != table) {
|
|
if (current) {
|
|
tm.set_tablet_map(current->table, std::move(current->map));
|
|
}
|
|
auto tablet_count = row.get_as<int>("tablet_count");
|
|
auto tmap = tablet_map(tablet_count);
|
|
current = active_tablet_map{table, tmap, tmap.first_tablet()};
|
|
}
|
|
|
|
tablet_replica_set tablet_replicas;
|
|
if (row.has("replicas")) {
|
|
tablet_replicas = deserialize_replica_set(row.get_view("replicas"));
|
|
}
|
|
|
|
tablet_replica_set new_tablet_replicas;
|
|
if (row.has("new_replicas")) {
|
|
new_tablet_replicas = deserialize_replica_set(row.get_view("new_replicas"));
|
|
}
|
|
|
|
if (!new_tablet_replicas.empty()) {
|
|
std::unordered_set<tablet_replica> pending(new_tablet_replicas.begin(), new_tablet_replicas.end());
|
|
for (auto&& r : tablet_replicas) {
|
|
pending.erase(r);
|
|
}
|
|
if (pending.size() > 1) {
|
|
throw std::runtime_error(format("Too many pending replicas for table {} tablet {}: {}",
|
|
table, current->tid, pending));
|
|
}
|
|
current->map.set_tablet_transition_info(current->tid, tablet_transition_info{
|
|
std::move(new_tablet_replicas), *pending.begin()});
|
|
}
|
|
|
|
current->map.set_tablet(current->tid, tablet_info{std::move(tablet_replicas)});
|
|
|
|
auto persisted_last_token = dht::token::from_int64(row.get_as<int64_t>("last_token"));
|
|
auto current_last_token = current->map.get_last_token(current->tid);
|
|
if (current_last_token != persisted_last_token) {
|
|
tablet_logger.debug("current tablet_map: {}", current->map);
|
|
throw std::runtime_error(format("last_token mismatch between on-disk ({}) and in-memory ({}) tablet map for table {} tablet {}",
|
|
persisted_last_token, current_last_token, table, current->tid));
|
|
}
|
|
|
|
current->tid = *current->map.next_tablet(current->tid);
|
|
};
|
|
|
|
try {
|
|
co_await qp.query_internal("select * from system.tablets",
|
|
[&] (const cql3::untyped_result_set_row& row) -> future<stop_iteration> {
|
|
process_row(row);
|
|
return make_ready_future<stop_iteration>(stop_iteration::no);
|
|
});
|
|
} catch (...) {
|
|
if (current) {
|
|
std::throw_with_nested(std::runtime_error(format("Failed to read tablet metadata for table {}", current->table)));
|
|
} else {
|
|
std::throw_with_nested(std::runtime_error("Failed to read tablet metadata"));
|
|
}
|
|
}
|
|
if (current) {
|
|
tm.set_tablet_map(current->table, std::move(current->map));
|
|
}
|
|
tablet_logger.trace("Read tablet metadata: {}", tm);
|
|
co_return std::move(tm);
|
|
}
|
|
|
|
future<std::vector<canonical_mutation>> read_tablet_mutations(seastar::sharded<service::storage_proxy>& proxy) {
|
|
auto s = db::system_keyspace::tablets();
|
|
auto rs = co_await db::system_keyspace::query_mutations(proxy, db::system_keyspace::NAME, db::system_keyspace::TABLETS);
|
|
std::vector<canonical_mutation> result;
|
|
result.reserve(rs->partitions().size());
|
|
for (auto& p: rs->partitions()) {
|
|
result.emplace_back(canonical_mutation(p.mut().unfreeze(s)));
|
|
}
|
|
co_return std::move(result);
|
|
}
|
|
|
|
}
|