Adds a heavy test which tests compatibility of BTI index files
between Cassandra and Scylla.
It's composed from a C++ part, used to read and write BTI files
with Scylla's readers and writers,
and a Python part which uses a Cassandra node and the C++
executable to make them read and write each other's files.
The stages of the test are:
1. Use the C++ part to generate a random BIG sstable, and matching
BTI index files.
2. Import the BIG files into Cassandra, let it generate its own
BTI index files.
3. Read both Scylla's BTI and Cassandra's BTI index files using
the C++ part. Check that they return the right positions
and tombstones for each partition and row.
4. Sneakily swap Cassandra's BTI files for Scylla's BTI files,
and query Cassandra (via CQL) for each row. Check that each query returns
the right result.
Not much can be inferred about the index via CQL queries,
so the check we are doing on Cassandra is relatively weak.
But in conjunction with the checks done on the Scylla part,
it's probably good enough.
The test is weird enough, and with heavy-enough dependencies
(it uses a podman container to run the Cassandra) that
ith has been put in test/manual.
To run the test, build
`build/$build_mode/test/manual/bti_cassandra_compatibility_test_g`,
and run `python test/manual/bti_cassandra_compatibility_test.py`.
Note: there's a lot of things that could go wrong in this test.
(E.g. file permission issues or port mapping issues due to the container
usage, incompatibilities between the Python driver and the random CQL
values generated by generate_random_mutations, etc).
I hope it works everywhere, but I only tested it on my machine,
running it inside the dbuild container.
We are moving away from integer generations, so stop using them.
Also drop the --generation command-line parameter, UUID generations
don't have be provided by the caller, because random UUIDs will not
collide with each other. To help the caller still know what generation
the output sstable has (previously they provided it via --generation),
print the generation to stdout.
Closesscylladb/scylladb#25166
We have an xfailing test test_secondary_index.py::test_limit_partition
which reproduces a Scylla bug in LIMIT when scanning a secondary index
(Refs #22158). The point of such a reproducer is to demonstrate the bug
by passing on Cassandra but failing on Scylla - yet this specific test
doesn't pass on Cassandra because it expects the wrong 3 out of 4
results to be returned:
The test begins with LIMIT 1 and sees the first result is (2,1), so we
expect when we increase the LIMIT to 3 to see more results from the same
partition (2) - and yet the test mistakenly expected the next results
to come from partition 1, which is not a reasonable expectation, and
doesn't happen in Cassandra (I checked both Cassandra 5 and 4).
After this patch, the test passes on Cassandra (I tried 4 and 5), and
continues to fail on Scylla - which returns 4 rows despite the LIMIT 3.
Note that it is debatable whether this test should insist at all
on which 3 items are returned by "LIMIT 3" - In Cassandra the ordering of
a SELECT with a secondary index is not well defined (see discussion in
Refs #23392). So an alternative implementation of this test would be
to just check that LIMIT 3 returns 3 items without insisting which:
# In Cassandra the ordering of a SELECT with a secondary index is not
# defined (see discussion in #23392), so we don't know which three
# results to expect - just that it must be a 3-item subset.
rows = list(rs)
assert len(rows) == 3
assert set(rows).issubset({(1,1), (1,2), (2,1), (2,2)})
However, as of yet, I did not modify this test to do this. I still
believe there is value in secondary index scans having the same order
as a scan without a secondary index has - and not an undefined order,
and if both Scylla and Cassandra implement that in practice, it's
useful for tests to validate this so we'll know if this guarantee is
ever broken.
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closesscylladb/scylladb#25676
The PRUNE MATERIALIZED VIEW statement is supposed to remove ghost rows from the
view. Ghost rows are rows in the view with no corresponding row in the base table.
Before this patch, only rows whose primary key columns of the base table had
different values than any of the base rows were treated as ghost rows by the PRUNE
statement. However, view rows which have a column in their primary key that's not
in the base primary can also be ghost rows if this column has a different value
than the base row with the same values of remaining primary key columns. That's
because these rows won't be deleted unless we change value of this column in the
base table to this specific value.
In this patch we add a check for this column in the PRUNE MATERIALIZED VIEW logic.
If this column isn't the same in the base table and the view, these rows are also
deleted.
Fixes https://github.com/scylladb/scylladb/issues/25655Closesscylladb/scylladb#25720
The test could trigger gossiper API calls before the API was properly registered, causing intermittent 404 errors. Previously the test waited for the "init - starting gossiper" log, but this appears before API registration completes.
Add explicit wait for gossiper API registration to ensure the endpoint is available before making requests, eliminating test flakiness.
Fixes: scylladb/scylladb#25582
No backport needed: Issue only observed in master so far.
Closesscylladb/scylladb#25583
* https://github.com/scylladb/scylladb:
test: improve async execution in test_long_join
test: fix race condition in test_long_join
In the Raft-based topology, a decommissioning node is removed from group
0 after the decommission request is considered finished (and the token
ring is updated). Therefore, `check_token_ring_and_group0_consistency`
called just after decommission might fail when the decommissioned node
is still in group 0 (as a non-voter). We deflake all tests that call
`check_token_ring_and_group0_consistency` after decommission in this
commit.
Fixes#25809
In the Raft-based topology, a decommissioning node is removed from group
0 after the decommission request is considered finished (and the token
ring is updated). `wait_for_token_ring_and_group0_consistency` doesn't
handle such a case; it only handles cases where the token ring is
updated later. We fix this in this commit.
We rely on the new implementation of
`wait_for_token_ring_and_group0_consistency` in the following commit to
fix flakiness of some tests.
We also update the obsolete docstring in this commit.
Consider the following:
1) balancer emits split decision
2) split compaction starts
3) split decision is revoked
4) emits merge decision
5) completes merge, before compaction in step 2 finishes
After last step, split compaction initiated in step 2 can fail because it works with the global tablet map, rather than the map when the compaction started. With the global state changing under its feet, on merge, the mutation splitting writer will think it's going backwards since sibling tablets are merged.
This problem was also seen when running load-and-stream, where split initiated by the sstable writer failed, split completed, and the unsplit sstable is left in the table dir, causing problems in the restart.
To fix this, let's make split compaction always work with the state when it started, not a global state.
Fixes#24153.
All 2025.* versions are vulnerable, so fix must be backported to them.
Closesscylladb/scylladb#25690
* github.com:scylladb/scylladb:
replica: Fix split compaction when tablet boundaries change
replica: Futurize split_compaction_options()
We modify the logic to make sure that all of the keyspaces that the test
creates are RF-rack-valid. For that, we distribute the nodes across two
DCs and as many racks as the provided replication factor.
That may have an effect on the load balancing logic, but since this is
a performance test and since tablet load balancing is still taking place,
it should be acceptable.
This commit also finishes work in adjusting perf tests to pass with
the `rf_rack_valid_keyspaces` configuration option enabled. The remaining
tests either don't attempt to create keyspaces or they already create
RF-rack-valid keyspaces.
We don't need to explicitly enable the configuration option. It's already
enabled by default by `cql_test_config`. The reason why we haven't run into
any issue because of that is that performance tests are not part of our CI.
Fixesscylladb/scylladb#25127Closesscylladb/scylladb#25728
Add another level of verbosity: quiet.
Before this it was used as a default one, but it provides not enough information.
These changes should be coupled with pytest-sugar plugin to have an intended information for each level.
Invoke the pytest as a module, instead of a separate process, to get access to the terminal to be able to it interactively.
Framework change only, so backporting in to 2025.3
Fixes: #25403Closesscylladb/scylladb#25698
* github.com:scylladb/scylladb:
test.py: add additional level of verbosity for output
test.py: start pytest as a module instead of subprocess
This patch introduces a new `incremental_mode` parameter to the tablet
repair REST API, providing more fine-grained control over the
incremental repair process.
Previously, incremental repair was on and could not be turned off. This
change allows users to select from three distinct modes:
- `regular`: This is the default mode. It performs a standard
incremental repair, processing only unrepaired sstables and skipping
those that are already repaired. The repair state (`repaired_at`,
`sstables_repaired_at`) is updated.
- `full`: This mode forces the repair to process all sstables, including
those that have been previously repaired. This is useful when a full
data validation is needed without disabling the incremental repair
feature. The repair state is updated.
- `disabled`: This mode completely disables the incremental repair logic
for the current repair operation. It behaves like a classic
(pre-incremental) repair, and it does not update any incremental
repair state (`repaired_at` in sstables or `sstables_repaired_at` in
the system.tablets table).
The implementation includes:
- Adding the `incremental_mode` parameter to the
`/storage_service/repair/tablet` API endpoint.
- Updating the internal repair logic to handle the different modes.
- Adding a new test case to verify the behavior of each mode.
- Updating the API documentation and developer documentation.
Fixes#25605Closesscylladb/scylladb#25693
Populate the local state during gossiper initialization in start_gossiping, preventing an empty state from being added to _endpoint_state_map and returned in get_endpoint_states responses, that was causing an 'empty host id issue' on the other nodes during nodes restart.
Check for a race condition in do_apply_state_locally In do_apply_state_locally, a race condition can occur if a task is suspended at a preemption point while the node entry is not locked.
During this time, the host may be removed from _endpoint_state_map. When the task resumes, this can lead to inserting an entry with an empty host ID into the map, causing various errors, including a node crash.
This change adds a check after locking the map entry: if a gossip ACK update does not contain a host ID, we verify that an entry with that host ID still exists in the gossiper’s _endpoint_state_map.
Fixes https://github.com/scylladb/scylladb/issues/25831
Fixes https://github.com/scylladb/scylladb/issues/25803
Fixes https://github.com/scylladb/scylladb/issues/25702
Fixes https://github.com/scylladb/scylladb/issues/25621
Ref https://github.com/scylladb/scylla-enterprise/issues/5613
Backport: The issue affects all current releases(2025.x), therefore this PR needs to be backported to all 2025.1-2025.3.
Closesscylladb/scylladb#25849
* github.com:scylladb/scylladb:
gossiper: fix empty initial local node state
gossiper: add test for a race condition in start_gossiping
gossiper: check for a race condition in `do_apply_state_locally`
test/gossiper: add reproducible test for race condition during node decommission
The test could trigger gossiper API calls before the API was properly
registered, causing intermittent 404 errors. Previously the test waited
for the "init - starting gossiper" log, but this appears before API
registration completes.
Add explicit wait for gossiper API registration to ensure the endpoint
is available before making requests, eliminating test flakiness.
Fixes: scylladb/scylladb#25582
This patch introduces one new test case. It tests that a keyspace
can be created without specifying replication options. Since other
options already had defaults, this test assures a keyspace can
be created with no options specified at all, with the following query:
`CREATE KEYSPACE ks;`
Refs #25145
This change removes the addition of an empty state to `_endpoint_state_map`.
Instead, a new state is created locally and then published via replicate,
avoiding the issue of an empty state existing in `_endpoint_state_map`
before the preemption point. Since this resolves the issue tested in
`test_gossiper_empty_self_id_on_shadow_round`, the `xfail` mark has been removed.
Fixes: scylladb/scylladb#25831
This change adds a test for a race condition in `start_gossiping` that
can lead to an empty self state sent in `gossip_get_endpoint_states_response`.
Test for scylladb/scylladb#25831
In do_apply_state_locally, a race condition can occur if a task is
suspended at a preemption point while the node entry is not locked.
During this time, the host may be removed from _endpoint_state_map.
When the task resumes, this can lead to inserting an entry with an
empty host ID into the map, causing various errors, including a node
crash.
This change
1. adds a check after locking the map entry: if a gossip ACK update
does not contain a host ID, we verify that an entry with that host ID
still exists in the gossiper’s _endpoint_state_map.
2. Removes xfail from the test_gossiper_race test since the issue is now
fixed.
3. Adds exception handling in `do_shadow_round` to skip responses from
nodes that sent an empty host ID.
This re-applies the commit 13392a40d4 that
was reverted in 46aa59fe49, after fixing
the issues that caused the CI to fail.
Fixes: scylladb/scylladb#25702Fixes: scylladb/scylladb#25621
Ref: scylladb/scylla-enterprise#5613
This change introduces a targeted test that simulates the gossiper race
condition observed during node decommissioning. The test delays gossip
state application and host ID lookup to reliably reproduce the scenario
where `gossiper::get_host_id()` is called on a removed endpoint,
potentially triggering an abort in `apply_new_states`.
There is a specific error injection added to widen the race window, in
order to increase the likelihood of hitting the race condition. The
error injection is designed to delay the application of gossip state
updates, for the specific node that is being decommissioned. This should
then result in the server abort in the gossiper.
This re-applies the commit 5dac4b38fb that
was reverted in dc44fca67c, but modified
to relax the check from "on_internal_error" to a just warning log. The
more strict can be re-introduced later once we are sure that all
remaining problems are resolved and it will not break the CI.
Refs: scylladb/scylladb#25621Fixes: scylladb/scylladb#25721
Sometimes `vector_store_client_test_ann_request` test hangs up. It is hard to
reproduce.
It seems that the problem is that tests are unreliable in case of stalled
requests. This patch attaches a timer to the abort_source to ensure that
the test will finish with a timeout at least.
Fixes: VECTOR-150
Fixes: #25234Closesscylladb/scylladb#25301
Consider the following scenario:
- Current replica set is [A, B, C]
- write succeeds on [A, B], and a hint is logged for node C
- before the hint is replayed, D bootstraps and the token migrates from C to D
- hint is replayed to node C while D is pending, but it's too late, since streaming for that token is already done
- C is cleaned up, replayed data is lost, and D has a stale copy until next repair.
In the scenario we effectively fail to send the hint. This scenario is also more likely to happen with tablets,
as it can happen for every tablet migration.
This issue is particularly detrimental to materialized views. View updates use hints by default and a specific
view update may be sent to just one view replica (when a single base replica has a different row state due to
reordering or missed writes). When we lose a hint for such a view update, we can generate a persistent inconsistency
between the base and view - ghost rows can appear due to a lost tombstone and rows may be missing in the view due
to a lost row update. Such inconsistencies can't be fixed neither by repairing the view or the base table.
To handle this, in this patch we add the pending replicas to the list of targets of each hint, even if the original
target is still alive.
This will cause some updates to be redundant. These updates are probably unavoidable for now, but they shouldn't
be too common either. The scenarios for them are:
1. managing to send the hint to the source of a migrating replica before streaming that its token - the write will
arrive on the pending replica anyway in streaming
2. the hint target not being the source of the migration - if we managed to apply the original write of the hint to
the actual source of the migration, the pending replica will get it during streaming
3. sending the same hint to many targets at a similar time - while sending to each target, we'll see the same pending
replica for the hint so we'll send it multiple times
4. possible retries where even though the hint was successfully sent to the main target, we failed to send it to the
pending replica, so we need to retry the entire write
This patch handles both tablet migrations and tablet rebuilds. In the future, for tablet migrations, we can avoid
sending the hint to pending replias if the hint target is not the source fo the migration, which would allow us to
avoid the redundant writes 2 and 3. For rack-aware RF, this will be as simple as checking whether the replicas are
in the same rack.
We also add a test case reproducing the issue.
Co-Authored-By: Raphael S. Carvalho <raphaelsc@scylladb.com>
Fixes https://github.com/scylladb/scylladb/issues/19835Closesscylladb/scylladb#25590
Interval's copy and move constructors are full of branches since the two payload T:s are
optional and therefore have to be optionally-constructed. This can be eliminated for
trivially copyable types (like dht::token) by eliminating interval's user-defined special member
functions (constructors etc) in that special case.
In turn, this enables optimizations in the standard library (and our own containers) that
convert moves/copies of spans of such types into memcpy().
Minor optimization, not a candidate to backport.
Closesscylladb/scylladb#25841
* github.com:scylladb/scylladb:
test: nonwrapping_interval_test: verify an interval of tokens is trivial
interval: specialize interval_data<T> for trivial types
interval: split data members into new interval_data class
Consider the following:
1) balancer emits split decision
2) split compaction starts
3) split decision is revoked
4) emits merge decision
5) completes merge, before compaction in step 2 finishes
After last step, split compaction initiated in step 2 can fail
because it works with the global tablet map, rather than the
map when the compaction started. With the global state changing
under its feet, on merge, the mutation splitting writer will
think it's going backwards since sibling tablets are merged.
This problem was also seen when running load-and-stream, where
split initiated by the sstable writer failed, split completed,
and the unsplit sstable is left in the table dir, causing
problems in the restart.
To fix this, let's make split compaction always work with
the state when it started, not a global state.
Fixes#24153.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
I want to write a test which generates a random table (random schema,
random data) and uses the Python driver to query it.
But it turns out that some values generated by test/lib/random_schema
can't be deserialized by the Python driver.
For example, it doesn't unknown uuid versions, dates before year 1
of after year 9999, or `time` values greater or equal to the number
of nanoseconds in a day.
AFAIK those "driver-illegal" values aren't particularly interesting
for tests which use `random_schema`, so we can just not generate
them.
Adds a fat unit test (integration test?) for BTI index writers and readers.
The test generates a small random dataset for the index writer, writes
it to a BTI file, and then attempts to run all possible (and legal)
sequences (up to a certain length) of index reader operations and check
that the results match the expectation (dictated by a "simple"
reference index reader).
It is currently the only test for this new feature, but I think it's reasonably
good. (I validated the coverage by looking at LLVM coverage reports
and by manually adding bugs in various places and checking that the test
catches it after a reasonably small number of runs).
(Note that in a later series, when we hook up BTI indexes
to Data.db readers/writers, the feature will also be indirectly
tested by the corresponding integration tests).
This is a randomized test. As such, its power grows with the number of runs.
In particular, a single run has a decently high probability of
not exercising parts of the code at all. (E.g. the generated dataset
might have no clustering keys).
Also this is a slow test. (With the default parameters,
~1s in release mode on my PC, several seconds in
debug mode. (And note that this is after a custom parameter
downsizing for debug mode, because this test is slowed extremely
badly by debug mode, due to the forced preemption after every future)).
For those two reasons, I'm not glad to include it in the test suite that runs in CI.
Instead of running it once in every CI run, I'd very much rather
have it run 10000 times after the tested feature is touched,
and before releases. Unfortunately we don't have a precedent for
something like that.
A `writer_node` contains a chain of multiple BTI nodes.
`writer_node::_node_size` is the size (in bytes) of the
entire chain.
But the parent of the `writer_node` wants to know the offset
to the rootmost node in the chain. This can be deduced from the
`writer_node::_transition_length` and the `writer_node::_node_size`.
But there's an error in the logic for that, for the special case
when there are two nodes in the chain.
The rootmost node will be SINGLE_NOPAYLOAD_12 if and only if
*the leafmost node* is smaller than 16 bytes, which is
true if and only if `_node_size` is smaller than 19 bytes.
But the current logic compares `_node_size` against 16.
That's incorrect. This patch fixes that.
There was a test for this branch, but it was not good enough.
It only tested payloads of size 1 and 20, but the problem
is only revealed by payloads of size 13-14.
The test has been extended to cover all sizes between 1 and 20.
As far as I know, positions outside the clustering region are never
passed to sstable indexes. But since they are representable within
the argument type of lazy_comparable_bytes_from_clustering_position,
let's make it handle them.
Before the patch, user with CREATE access could create a table with CDC or alter the table enabling CDC, but could not query a SELECT on the CDC table they created.
It was due to the fact, the SELECT permission was checked on the CDC log, and later it's "parent" - the keyspace, but not the base table, on which the user had SELECT permission automatically granted on CREATE.
This patch matches the behavior of querying the CDC log to the one implemented for Materialized Views:
1. No new permissions are granted on CREATE.
2. When querying SELECT, the permissions on base table SELECT are checked.
Fixes: https://github.com/scylladb/scylladb/issues/19798
Fixes: VECTOR-151
Closes scylladb/scylladb#25797
* github.com:scylladb/scylladb:
cqlpy/test_permissions: run the reproducer tests for #19798
select_statement: check for access to CDC base table
The branch inside the `if constexpr (debug)` contains a piece
of template code that doesn't typecheck properly.
(I used this code before committing it, but apparently
I let it become outdated when some changes around it
happened).
Fix that.
This patch adds LSI tests that correspond to already existing tests for
GSI, filling up some gaps in LSI:
- `_*wrong_type_attribute_*`,
- `test_lsi_duplicate_with_different_name`,
- `test_lsi_empty_value_in_batch_write`,
- `test_lsi_null_index`.
As well as some new ones:
- `test_lsi_empty_value_binary`,
- `test_lsi_put_overwrites_lsi_column`,
- `test_lsi_*_modifies_index`.
Before this change, executing `DESCRIBE MATERIALIZED VIEW` on the underlying
materialized view of a secondary index would produce a `CREATE INDEX` statement.
It was not only confusing, but it also prevented from learning about
the definition of the view. The only way to do so was to query system tables.
We change that behavior and produce a `CREATE MATERIALIZED VIEW` statement
instead. The statement is printed as a comment to implicitly convey that
the user should not attempt to execute it to restore the view. A short comment
is provided to make it clearer.
Before this commit:
```
cqlsh> CREATE TABLE ks.t(p int PRIMARY KEY, v int);
cqlsh> CREATE INDEX i ON ks.t(v);
cqlsh> DESCRIBE MATERIALIZED VIEW ks.i;
CREATE INDEX i ON ks.t(v);
```
After this commit:
```
cqlsh> CREATE TABLE ks.t(p int PRIMARY KEY, v int);
cqlsh> CREATE INDEX i ON ks.t(v);
cqlsh> DESCRIBE MATERIALIZED VIEW ks.i;
/* Do NOT execute this statement! It's only for informational purposes.
This materialized view is the underlying materialized view of a secondary
index. It can be restored via restoring the index.
CREATE MATERIALIZED VIEW ks.i_index [...];
*/
```
Note that describing the base table has not been affected and still works
as follows:
```
cqlsh> CREATE TABLE ks.t(p int PRIMARY KEY, v int);
cqlsh> CREATE INDEX i ON ks.t(v);
cqlsh> DESCRIBE TABLE ks.t;
CREATE TABLE ks.t (
p int,
v int,
PRIMARY KEY (p)
) WITH bloom_filter_fp_chance = 0.01
AND caching = {'keys': 'ALL', 'rows_per_partition': 'ALL'}
AND comment = ''
AND compaction = {'class': 'IncrementalCompactionStrategy'}
AND compression = {'sstable_compression': 'org.apache.cassandra.io.compress.LZ4Compressor'}
AND crc_check_chance = 1
AND default_time_to_live = 0
AND gc_grace_seconds = 864000
AND max_index_interval = 2048
AND memtable_flush_period_in_ms = 0
AND min_index_interval = 128
AND speculative_retry = '99.0PERCENTILE'
AND tombstone_gc = {'mode': 'timeout', 'propagation_delay_in_seconds': '3600'};
CREATE INDEX i ON ks.t(v);
```
We also provide two reproducers of scylladb/scylladb#24610.
Fixesscylladb/scylladb#24610Closesscylladb/scylladb#25697
Determine the progress of compaction tasks that have
children.
The progress of a compaction task is calculated using the default
get_progress method. If the expected_total_workload method is
implemented, the default progress is computed as:
(sum of child task progresses) / (expected total workload)
If expected_total_workload is not defined, progress is estimated based
on children progresses. However, in this case, the total progress may
increase over time as the task executes.
All compaction tasks, except for reshape tasks, implement the
expected_children_number method. To compute expected_total_workload,
iterate over all SSTables covered by the task and sum their sizes. Note
that expected_total_workload is just an approximation and the real workload
may differ if SStables set for the keyspace/table/compaction group changes.
Reshape tasks are an exception, as their scope is determined during
execution. Hence, for these tasks expected_total_workload isn't defined
and their progress (both total and completed) is determined based
on currently created children.
Fixes: https://github.com/scylladb/scylladb/issues/8392.
Fixes: https://github.com/scylladb/scylladb/issues/6406.
Fixes: https://github.com/scylladb/scylladb/issues/7845.
New feature, no backport needed
Closesscylladb/scylladb#15158
* github.com:scylladb/scylladb:
test: add compaction task progress test
compaction: set progress unit for compaction tasks
compaction: find expected workload for reshard tasks
compaction: find expected workload for global cleanup compaction tasks
compaction: find expected workload for global major compaction tasks
compaction: find expected workload for keyspace compaction tasks
compaction: find expected workload for shard compaction tasks
compaction: find expected workload for table compaction tasks
compaction: return empty progress when compaction_size isn't set
compaction: update compaction_data::compaction_size at once
tasks: do not check expected workload for done task
When creating a new keyspace, replication factor must be stated.
For example:
`CREATE KEYSPACE ks WITH REPLICATION { 'class': 'NetworkTopologyStrategy', 'replication_factor': 3 };`
This patch changes it in the following way - if there is no
replication factor specified, use default replication factor.
Default replication factor is equal to the number of racks that
are not arbiter-only, i.e. racks that have at least one non-arbiter node.
The following syntax is now valid:
`CREATE KEYSPACE ks WITH REPLICATION { 'class': 'NetworkTopologyStrategy' };`
`CREATE KEYSPACE ks WITH REPLICATION { };`
Fixes#16028
Backport is not needed. This is an enhancement for future releases.
Closesscylladb/scylladb#25570
* github.com:scylladb/scylladb:
docs/cql: update documentation for default replication factor
test/cqlpy: add keyspace creation default replication factor tests
cql3: add default replication factor to `create_keyspace_statement`
Fixes#25683
Once a table drop is complete, there should be no reason to retain
truncation records for it, as any replay should skip mutations
anyway (no CF), and iff we somehow resurrect a dropped table,
this replay-resurrected data is the least problem anyway.
Adds a prune phase to the startup drop_truncation_rp_records run,
which ignores updating, and instead deletes records for non-existant
tables (which should patch any existing servers with lingering data
as well).
Also does an explicit delete of records on actual table DROP, to
ensure we don't grow this table more than needed even in long
uptime nodes.
Small unit test included.
Closesscylladb/scylladb#25699
In do_apply_state_locally, a race condition can occur if a task is
suspended at a preemption point while the node entry is not locked.
During this time, the host may be removed from _endpoint_state_map.
When the task resumes, this can lead to inserting an entry with an
empty host ID into the map, causing various errors, including a node
crash.
This change adds a check after locking the map entry: if a gossip ACK update
does not contain a host ID, we verify that an entry with that host ID
still exists in the gossiper’s _endpoint_state_map.
Fixesscylladb/scylladb#25702Fixesscylladb/scylladb#25621
Ref scylladb/scylla-enterprise#5613Closesscylladb/scylladb#25727
Adds infrastructure and client for interaction with GCP object storage services.
Note: this is just a client object usable for creating, listing, deleting and up/downloading of objects to/from said storage service. It makes no attempt at actually inserting it into the sstable storage flow. That can come later.
This PR breaks out GCP auth and some general REST call functionality into shared routines. Not all code is 100% reused, but at least some.
Test is added, though could be more comprehensive (feel free to suggest a test vector).
Test can run in either local mock server mode (default), or against actual GCP.
See `test/boost/gcp_object_storage_test.cc` for explanation on the config environment vars.
Default is to run the test against a temporary docker deamon.
Closesscylladb/scylladb#24629
* github.com:scylladb/scylladb:
test::boost::gcp_object_storage_test: Initial unit tests for GCP obj storage
proc-utils: Re-export waiting types from seastar
proc-utils: Inherit environment from current process
utils::gcp::object_storage: Add client for GCP object storage
utils::http: Add optional external credentials to dns_connection_factory init
utils::rest: Break out request wrapper and send logic
encryption::gcp_host: Use shared gcp credentials + REST helpers
utils::gcp: Move/add gcp credentials management to shared file
utils::rest::client: Add formatter for seastar::http::reply
utils::rest::client: Add helper routines for simple REST calls
utils::http: Make shared system trust certificates public
Normally, when we create a table, MV, etc., we apply `cf_prop_defs` to the
schema builder via the function `cf_prop_defs::apply_to_builder`. Unfortunately,
that didn't happen when creating CDC log tables, and so we might have missed
some of the properties that would normally be set to some value, even if the
default one.
One particular example of that phenomenon was `tombstone_gc`. For better or
worse, it's not a "standalone property" of a table, but rather part of
`extensions`. [Somewhat related issue: scylladb/scylladb#9722]
That may have and did cause trouble. Consider this scenario:
1. A CDC log table is created.
2. The table does NOT have any value of `tombstone_gc` set.
3. The user edits the table via `ALTER TABLE`. That statement treats the log
table just like any other one (at least as far as the relevant portion of the
logic is concerned). Among other things, it uses
`cf_prop_defs::apply_to_builder`, and as a result, the `tombstone_gc`
property is set to some value:
* the default one if the user doesn't specify it in the statement,
* a custom one if they do.
Why is that a problem?
First of all, it's confusing. When we perform a schema backup and a table uses
CDC, we include an ALTER statement for its corresponding CDC log table (for more
context, see issue scylladb/scylladb#18467 or commit
scylladb/scylladb@f12edbdd95).
There are two consequences for the user here:
1. If the log table had NOT been altered ever since it was created, the
statement will miss the `tombstone_gc` property as if it couldn't be set for
it at all. That's confusing!
2. If the log table HAD in fact been altered after its creation, the statement
will include the `tombstone_gc` property. That's even more confusing (why was
it not present the first time, but it is now?).
The `tombstone_gc` property should always be set to avoid confusion and
problematic edge cases in tests and to simply be consistent with how other
schema entities work.
The solution we employ is that we always set the property to the default
value. That includes the case when we reattach the log table to the base;
consider the following scenario:
1. Create a table with CDC enabled.
2. Detach the log table by performing `ALTER TABLE ... WITH cdc = {'enabled': false}`.
3. Change the `tombstone_gc` property of the log table.
4. Reattach the log table to the base in the same way as in step 2.
The expected result would be that the new value of `tombstone_gc` would be
preserved after reattaching the log table. However, that's not what will
happen. We decide to stay consistent with how other properties of a log
table behave, and we reset them after every reattachment. We might change that
in the future: see issue scylladb/scylladb#25523.
Two reproducer tests of scylladb/scylladb#25187 are included in the changes.
Backport: The problem is not critical, so it may not be necessary to backport the changes.
That's to be discussed.
Closesscylladb/scylladb#25521
* github.com:scylladb/scylladb:
cdc: Set tombstone_gc when creating log table
tombstone_gc: Add overload of get_default_tombstone_gc_mode
tombstone_gc: Rename get_default_tombstonesonte_gc_mode
Consider the following scenario:
- A tablet is migrated away from a shard
- The tablet cleanup stage closes the storage group's async_gate
- A drop table runs truncate which attempts to disable compaction on the tablet with its gate closed. This fails, because table::parallel_foreach_compaction_group() ultimately calls storage_group_manager::parallel_foreach_storage_group() which will not disable compaction if it can't hold the storage group's gate
- Truncate calls table::discard_sstables() which checks if the compaction has been disabled, and because it hasn't, it then runs on_internal_error() with "compaction not disabled on table ks.cf during TRUNCATE" which causes a crash
Fixes: #25706
This needs to be backported to all supported versions with tablets
Closesscylladb/scylladb#25708
* github.com:scylladb/scylladb:
test: reproducer and test for drop with concurrent cleanup
truncate: check for closed storage group's gate in discard_sstables
Allows testing using either local mock server (installed or using docker),
or real GCP project (not tested as of writing this).
v2: Try podman if docker unavail
v3: Ensure we check log output on fake-gcs, because when using podman, the
published port will be connectible even though the actual server is not
up yet.
v4: Use ephermal port forward in docker/podman to allow us running parallel
instances. Also adjust credentials and port finding in test.
v5: Re-ensure no parallel tests for this: We seem to time out in podman
trying to fetch image for X parallel tests
v6: Remove the ephermal port stuff. Because of course this does not work
with our podman-in-podman. Do brute-force port speculation instead.
v7: Up timeout for server start to allow docker pull.
v8: Fix string check error
v9: Add explicit docker image version
Executing a vector search (SELECT with ANN OF ordering) query with `TRACING ON` enabled
caused a node to crash due to a null pointer dereference.
This occurred because a vector index does not have an associated view
table, making its `_view_schema` member null. The implementation
attempted to enable tracing on this null view schema, leading to the
crash.
The fix adds a null check for `_view_schema` before attempting to
enable tracing on the view (index) table.
A regression test is included to prevent this from happening again.
Fixes: VECTOR-179
Closesscylladb/scylladb#25500
