This PR adds support for limiting the maximum shares allocated to a
compaction scheduling class by the compaction controller. It introduces
a new configuration parameter, compaction_max_shares, which, when set
to a non zero value, will cap the shares allocated to compaction jobs.
This PR also exposes the shares computed by the compaction controller
via metrics, for observability purposes.
Fixes https://github.com/scylladb/scylladb/issues/9431
Enhancement. No need to backport.
NOTE: Replaces PR https://github.com/scylladb/scylladb/pull/26696
Ran a test in which the backlog raised the need for max shares (normalized backlog above normalization_factor), and played with different values for new option compaction_max_shares to see it works (500, 1000, 2000, 250, 50)
Closesscylladb/scylladb#27024
* github.com:scylladb/scylladb:
db/config: introduce new config parameter `compaction_max_shares`
compaction_manager:config: introduce max_shares
compaction_controller: add configurable maximum shares
compaction_controller: introduce `set_max_shares()`
There is a limited number of ways to obtain the schema of a table:
1) Use DESCRIBE TABLE in cqlsh
2) Find the schema definition in the code (for system tables)
3) Ask support/user to provide schema
4) Piece together the schema definition from the system tables
Option (1) is the most convenient but requires access to live cluster.
(2) is limited to system tables only.
When investigating issues for customers, we have to rely on (3) and this
often adds communication round-trips and delays. (4) requires knowledge
of ScyllaDB internals and access to system tables.
The new dump-schema commands provides a convenient way to obtain the
schema of tables, given that there is access to either an sstable or the
system tables. It can dump the schema of system tables without either.
Closesscylladb/scylladb#26433
With the introduction of colocated tables, all the tablet transitions
now operate on groups of colocated tablets instead of individual
tablets. such is tablet migration, and also tablet repair.
The tablet repair currently doesn't work on individual tablets due to
the limitations in the tablet map being shared. The way it was
implemented to work on a group of colocated tablets is by repairing all
the colocated tablets together, using a dedicated rpc, and setting a
shared repair_time in the shared tablet map. It was implemented this
way because we wanted to have some way to repair the tablets of a
colocated table.
However, we want to change this in the next release so that it will be
possible to repair the tablets of a colocated table individually. In
order to simplify and prepare for the future change, we prefer until
then to not repair colocated tables at all. otherwise, we will need to
support both the shared repair and individual repair together for a long
time, and the upgrade will be more complicated.
We change the handling of the tablet 'repair' transition to repair only
the base table's tablets. It means it will not be possible to request
tablet repair for a non-base colocated table such as local MV, CDC and
paxos table. This restriction will be temporary until a later release
where we will suuport repairing colocated tablets.
This is a reasonable restriction because repair for these kind of tables
is not required or as important as for normal tables.
Fixes https://github.com/scylladb/scylladb/issues/27119
backport to 2025.4 since we must change it in the same version it's introduced before it's released
Closesscylladb/scylladb#27120
* github.com:scylladb/scylladb:
tombstone_gc: don't use 'repair' mode for colocated tables
Revert "storage service: add repair colocated tablets rpc"
topology_coordinator: don't repair colocated tablets
The vector-search-validator is a binary tool which do functional and integration tests between scylla and vector-store. It is build in Rust mainly in vector-store repository. This patch adds possibility to write tests on scylladb repository side, compile them together with vector-store tests and run them in `test.py` environment.
There are three parts of the change:
- add sources of validator to the `test/vector_search_validator` directory
- add support for building validator and vector-store in `build/vector-search-validator/bin` directory with or without cmake
- add support for `pytest` and `test.py` to run validator test locally and in the CI environment; this part adds also README to the `test/vector_search_validator` directory
Design for validator integration tests: https://scylladb.atlassian.net/wiki/spaces/RND/pages/39518215/Vector+Search+Core+Test+Plan+Document
References: VECTOR-50
No backport needed as this is a new functionality.
Closesscylladb/scylladb#26653
* github.com:scylladb/scylladb:
vector_search: add vector-search-validator tests
vector_search: implement building vector-search-validator
vector_search: add vector-search-validator sources
For tables of special types that can be located: MV, CDC, and paxos
table, we should not use tombstone_gc=repair mode because colocated
tablets are never repaired, hence they will not have repair_time set and
will never be GC'd using 'repair' mode.
With the introduction of colocated tables, all the tablet transitions
now operate on groups of colocated tablets instead of individual
tablets. such is tablet migration, and also tablet repair.
The tablet repair currently doesn't work on individual tablets due to
the limitations in the tablet map being shared. The way it was
implemented to work on a group of colocated tablets is by repairing all
the colocated tablets together, using a dedicated rpc, and setting a
shared repair_time in the shared tablet map. It was implemented this
way because we wanted to have some way to repair the tablets of a
colocated table.
However, we want to change this in the next release so that it will be
possible to repair the tablets of a colocated table individually. In
order to simplify and prepare for the future change, we prefer until
then to not repair colocated tables at all. otherwise, we will need to
support both the shared repair and individual repair together for a long
time, and the upgrade will be more complicated.
We change the handling of the tablet 'repair' transition to repair only
the base table's tablets. It means it will not be possible to request
tablet repair for a non-base colocated table such as local MV, CDC and
paxos table. This restriction will be temporary until a later release
where we will suuport repairing colocated tablets.
This is a reasonable restriction because repair for these kind of tables
is not required or as important as for normal tables.
Fixesscylladb/scylladb#27119
Vector indexes are going to be supported only for tablets (see VECTOR-322).
As a result, tests using vector indexes will be failing when run with vnodes.
This change ensures tests using vector indexes run exclusively with tablets.
Fixes: VECTOR-49
Closesscylladb/scylladb#26843
The commit adds a functionality for `pytest` and `test.py` to run
`vector-search-validator` in `sudo unshare` environment. There are already two
tests - first parametrized `test_validator.py::test_validator[test-case-name]`
(run validator) and second `test_cargo_toml.py::test_cargo_toml` (check if the
current `Cargo.toml` for validator is correct).
Documentation for these tests are provided in `README.md`.
The commit adds targets building
`build/vector-search-validator/bin/{vector-store,vector-search-validator}. The
targets must be build for tests. They don't depend on build mode.
The commit adds target in `configure.py` and also in `cmake`.
The commit adds validator sources uses combination of local files and
vector-store's files. In `build-env` there are definition of vector-store git
repository and revision on which validator will be built. `cargo-toml-template`
is script for printing current `Cargo.toml` to the stdout. After updating
`build-env` developer needs to update new configuration with
`./cargo-toml-template > Cargo.toml`. Git revision is used in several places in
`Cargo.toml` and will be used for building `vector-store`, so for better
handling git revision it should be setup only in one place.
The validator is divided into several crates to be able to built it within
scylladb and vector-store repositories. Here we need to create a new validator
crate with simple `main` function and call `validator_engine::main` there. We
provide tests written in scylladb repo in `validator-scylla` crate. The commit
provides empty `cql` test case, which should be filled in the future.
Currently if a banned node tries to connect to a cluster it fails to
create connections, but has no idea why, so from inside the node it
looks like it has communication problems. This patch adds new rpc
NOTIFY_BANNED which is sent back to the node when its connection is
dropped. On receiving the rpc the node isolates itself and print an
informative message about why it did so.
Closesscylladb/scylladb#26943
Add support for the new configuration parameter `compaction_max_shares`,
and update the compaction manager to pass it down to the compaction
controller when it changes. The shares allocated to compaction jobs will
be limited by this new parameter.
Fixes#9431
Signed-off-by: Lakshmi Narayanan Sreethar <lakshmi.sreethar@scylladb.com>
Introduce an updateable value `max_shares` to compaction manager's
config. Also add a method `update_max_shares()` that applies the latest
`max_shares` value to the compaction controller’s `max_shares`. This new
variable will be connected to a config parameter in the next patch.
Refs #9431
Signed-off-by: Lakshmi Narayanan Sreethar <lakshmi.sreethar@scylladb.com>
When migrating tablet size during the end_migration tablet transition stage, we need the pending and leaving replica hosts. The leaving and pending replicas are gathered in objects of type std::optional<tablet_replica> and are not checked if they contain a value before dereferencing which could cause an exception in the topology coordinator.
This patch adds a check for leaving and pending replicas, and only performs the tablet size migration if neither are empty.
This bug was introduced in 10f07fb95a
This change also adds the ability to create a tablet size in load_stats during end_migration stage of a tablet rebuild. We compute the new tablet size from by averaging the tablet sizes of the existing replicas.
This change also adds the virtual table tablet_sizes which contains tablet sizes of all the replicas of all the tablets in the cluster.
A version containing this bug has not yet been released, so a backport is not needed.
Closesscylladb/scylladb#27118
* github.com:scylladb/scylladb:
test: add tests for tablet size migration during end_migration
virtual_table: add tablet_sizes virtual table
load_stats: update tablet sizes after migration or rebuild
This patch enables integrity check in 'create_stream_sources()' by introducing a new 'sstable_data_stream_source_impl' class for handling the Data component of SSTables. The new implementation uses 'sstable::data_stream()' with 'integrity_check::yes' instead of the raw input_stream.
These additional checks require reading the digest and CRC components from disk, which may introduce some I/O overhead. For uncompressed SSTables, this involves loading and computing checksums and digest from the data.
For compressed SSTables - where checksums are already embedded - the cost comes from reading, calculating and verifying the diges.
New test cases were added to verify that the integrity checks work correctly, detecting both data and digest mismatches.
Backport is not required, since it is a new feature
Fixes#21776Closesscylladb/scylladb#26702
* github.com:scylladb/scylladb:
file_stream_test: add sstable file streaming integrity verification test cases
streaming: prioritize sender-side errors in tablet_stream_files
sstables: enable integrity check for data file streaming
sstables: Add compressed raw streaming support
sstables: Allow to read digest and checksum from user provided file instance
sstables: add overload of data_stream() to accept custom file_input_stream_options
In ALTER KEYSPACE, when a datacenter name is omitted, its replication
factor is implicitly set to zero with vnodes, while with tablets,
it remains unchanged.
ALTER KEYSPACE should behave the same way for tablets as it does
for vnodes. However, this can be dangerous as we may mistakenly
drop the whole datacenter.
Reject ALTER KEYSPACE if it changes replication factor, but omits
a datacenter that currently contains tablet replicas.
Fixes: https://github.com/scylladb/scylladb/issues/25549.
Closesscylladb/scylladb#25731
Primary issue with the old method is that each update is a separate
cross-shard call, and all later updates queue behind it. If one of the
shards has high latency for such calls, the queue may accumulate and
system will appear unresponsive for mapping changes on non-zero shards.
This happened in the field when one of the shards was overloaded with
sstables and compaction work, which caused frequent stalls which
delayed polling for ~100ms. A queue of 3k address updates
accumulated, because we update mapping on each change of gossip
states. This made bootstrap impossible because nodes couldn't
learn about the IP mapping for the bootstrapping node and streaming
failed.
To protect against that, use a more efficient method of replication
which requires a single cross-shard call to replicate all prior
updates.
It is also more reliable, if replication fails transiently for some
reason, we don't give up and fail all later updates.
Fixes#26865Closesscylladb/scylladb#26941
* github.com:scylladb/scylladb:
address_map: Use barrier() to wait for replication
address_map: Use more efficient and reliable replication method
utils: Introduce helper for replicated data structures
Trie-based sstable indexes are supposed to be (hopefully) a better default than the old BIG indexes.
Make them the new default.
If we change our mind, this change can be reverted later.
New functionality, and this is a drastic change. No backport needed.
Closesscylladb/scylladb#26377
* github.com:scylladb/scylladb:
db/config: enable `ms` sstable format by default
cluster/dtest/bypass_cache_test: switch from highest_supported_sstable_format to chosen_sstable_format
api/system: add /system/chosen_sstable_version
test/cluster/dtest: reduce num_tokens to 16
This commit introduces TLS encryption support for vector store connections.
A new configuration option is added:
- vector_store_encryption_options.truststore: path to the trust store file
To enable secure connections, use the https:// scheme in the
vector_store_primary_uri/vector_store_secondary_uri configuration options.
Fixes: VECTOR-327
This change adds tests for the correctness of tablet size migration
during the end_migrations stage. This size migration can happend for
tablet migrations and for tablet rebuild.
Add 'test_sstable_stream' to verify SSTable file streaming integrity check.
The new tests cover both compressed and uncompressed SSTables and includes:
- Checksum mismatch detection verification
- Digest mismatch detection verifivation
Trie-based sstable indexes are supposed to be (hopefully)
a better default than the old BIG indexes.
Make them the new default.
If we change our mind, this change can be reverted later.
Trie-based indexes and older indexes have a difference in metrics,
and the test uses the metrics to check for bypass cache.
To choose the right metrics, it uses highest_supported_sstable_format,
which is inappropriate, because the sstable format chosen for writes
by Scylla might be different than highest_supported_sstable_format.
Use chosen_sstable_format instead.
Returns the sstable version currently chosen for use in for new sstables.
We are adding it because some tests want to know what format they are
writing (tests using upgradesstable, tests which check stats that only
apply to one of the index types, etc).
(Currently they are using `highest_supported_sstable_format` for this
purpose, which is inappropriate, and will become invalid if a non-latest
format is the default).
Add a test which verifies that if two nodes have the same data, with
different timestamps, repair will detect and fix the diverging
timestamps.
All our repair tests focus on difference in data and I remember writing
this test multiple times in the past to quickly verify whether this
works. Time to upstream this test.
Closesscylladb/scylladb#26900
DynamoDB's documentation
https://docs.aws.amazon.com/sdkref/latest/guide/feature-compression.html
suggests that DynamoDB allows request bodies to be compressed (currently
only by gzip).
The purpose of patch is to have a test reproducing this feature. The
test shows us that indeed DynamoDB understands compressed requests using
the "gzip" encoding, but Alternator does *not*, so the new test is xfail.
As you can see in the test code, although the low-level SDK (botocore)
can send compress requests, this is not actually enabled for DynamoDB
and we need to resort to some trickery to send compressed requests.
But the point is that once we do manage to send compressed requests,
the test shows us that they work properly on AWS, but fail on
Alternator.
The failure of the compressed requests on Alternator is reported like:
An error occurred (ValidationException) when calling the PutItem
operation: Parsing JSON failed: Invalid value. at 70459088
This error message should probably be improved (what is that high
number?!) but of course even better would be to make it really work.
By enabling tracing on alternator-server (e.g., edit test/cqlpy/run.py
and add `'--logger-log-level', 'alternator-server=trace',`) we can
see exactly what request the SDK sends Alternator. What we can see in
the request is:
1. The request headers are uncompressed (this is expected in HTTP)
2. There is a header "Content-Encoding: gzip"
3. The request's body is binary, a full-fleged gzip output complete with
a gzip magic in the beginning.
Refs #5041
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closesscylladb/scylladb#27049
Currently sstables_manager keeps a reference on global db::config to configure itself. Most of other services use their own specific configs with much less data on-board for the same purposes (e.g. #24841, #19051 and #23705 did same for other services) This PR applies this approach to sstables_manager as well.
Mostly it moves various values from db::config onto newly introduced struct sstables_manager::config, but it also adds specific tracking of sstable_file_io_extensions and patches tools/scylla-sstable not to use sstables_manager as "proxy" object to get db::config from along its calls.
Shuffling components dependencies, no need to backport
Closesscylladb/scylladb#27021
* github.com:scylladb/scylladb:
sstables_manager: Drop db::config from sstables_manager
tools/sstable: Make shard_of_with_tablets use db::config argument
tools/sstable: Add db::config& to all operations
tools/sstable: Get endpoints from storage manager
sstables_manager: Hold sstable IO extensions on it
sstables: Manager helper to grab file io extensions
sstables_manager: Move default format on config
sstables_manager: Move enable_sstable_data_integrity_check on config
sstables_manager: Move data_file_directories on config
sstables_manager: Move components_memory_reclaim_threshold on config
sstables_manager: Move column_index_auto_scale_threshold on config
sstables_manager: Move column_index_size on config
sstables_manager: Move sstable_summary_ratio on config
sstables_manager: Move enable_sstable_key_validation on config
sstables_manager: Move available_memory on config
code: Introduce sstables_manager::config
sstables: Patch get_local_directories() to work on vector of paths
code: Rename sstables_manager::config() into db_config()
When we enable CDC on a table, Scylla creates a log table for it.
It has default properties, but the user may change them later on.
Furthermore, it's possible to detach that log table by simply
disabling CDC on the base table:
```cql
/* Create a table with CDC enabled. The log table is created. */
CREATE TABLE ks.t (pk int PRIMARY KEY) WITH cdc = {'enabled': true};
/* Detach the log table. */
ALTER TABLE ks.t WITH cdc = {'enabled': false};
/* Modify a property of the log table. */
ALTER TABLE ks.t_scylla_cdc_log WITH bloom_filter_fp_chance = 0.13;
```
The log table can also be reattached by enabling CDC on the base table
again:
```cql
/* Reattach the log table */
ALTER TABLE ks.t WITH cdc = {'enabled': true};
```
However, because the process of reattachment goes through the same
code that created it in the first place, the properties of the log
table are rolled back to their default values. This may be confusing
to the user and, if unnoticed, also have other consequences, e.g.
affecting performance.
To prevent that, we ensure that the properties are preserved.
A reproducer test,
`test_log_table_preserves_properties_after_reattachment`, has been
provided to verify that the changes are correct.
Another test, `test_log_table_preserves_id_after_reattachment`, has
also been added because the current implementation sets properties
and the ID separately.
Fixesscylladb/scylladb#25523
Backport: not necessary. Although the behavior may be unexpected,
it's not a bug per se.
Closesscylladb/scylladb#26443
* github.com:scylladb/scylladb:
cdc: Preserve properties when reattaching log table
cdc: Extract creating columns in CDC log table to dedicated function
cdc: Extract default properties of CDC log tables to dedicated function
schema/schema_builder.hh: Add set_properties
schema: Add getter for schema::user_properties
schema: Remove underscores in fields of schema::user_properties
schema: Extract user properties out of raw_schema
Fixes#26874
Due to certain people (me) not being able to tell forward from backward,
the data alignment to ensure partial uploads adhere to the 256k-align
rule would potentially _reorder_ trailing buffers generated iff the
source buffers input into the sink are small enough. Which, as a fun fact,
they are in backup upload.
Change the unit test to use raw sink IO and add two unit tests (of which
the smaller size provokes the bug) that checks the same 64k buf segmented
upload backup uses.
Closesscylladb/scylladb#26938
Add a table name to Alternator's tracing output, as some clients would
like to consistently receive this information.
- add missing `tracing::add_table_name` in `executor::scan`
- add emiting tables' names in `trace_state::build_parameters_map`
- update tests, so when tracing is looked for it is filtered by table's
name, which confirms table is being outputed.
- change `struct one_session_records` declaration to `class one_session_records`,
as `one_session_records` is later defined as class.
Refs #26618Fixes#24031Closesscylladb/scylladb#26634
cluster.dtest_alternator_tests.test_slow_query_logging performs
a bootstrap with 768 token ranges.
It works with `me` sstables, which have 2 open file descriptors
per open sstable, but with `ms` sstables, which have 3 open
file descriptors per open sstable, it fails with EMFILE.
To avoid this problem, let's just decrease the number of vnodes
for in the test suite. It's appropriate anyway, because it avoids some
unneeded work without weakening the tests.
(Note: pylib-based have been setting `num_tokens` to 16 for a long time too).
This breaks `bypass_cache_test`, which is written in a way that expects
a certain number of token ranges. We adjust the relevant parameter
accordingly.
Consider the following:
1) single-key read starts, blocks on replica e.g. waiting for memory.
2) the same replica is migrated away
3) single-key read expires, coordinator abandons it, releases erm.
4) migration advances to cleanup stage, barrier doesn't wait on
timed-out read
5) compaction group of the replica is deallocated on cleanup
6) that single-key resumes, but doesn't find sstable set (post cleanup)
7) with abort-on-internal-error turned on, node crashes
It's fine for abandoned (= timed out) reads to fail, since the
coordinator is gone.
For active reads (non timed out), the barrier will wait for them
since their coordinator holds erm.
This solution consists of failing reads which underlying tablet
replica has been cleaned up, by just converting internal error
to plain exception.
Fixes#26229.
Signed-off-by: Raphael S. Carvalho <raphaelsc@scylladb.com>
Closesscylladb/scylladb#27078
According to the IETF spec uuid variant bits should be set to '10'. All
others are either invalid or reserved. The patch change the code to
follow the spec.
Closesscylladb/scylladb#27073
One of the tests in test_describe.py used "USE {test_keyspace}" which
affects the CQL session shared by all tests in an unrecoverable way (there
is no "UNUSE" statement).
As an example of what might happen if the shared CQL session is "polluted"
by a USE, issue #26334 is about a bug we have in DESC KEYSPACES when USE
is active.
So in this patch, we:
1. Fix the test to not use USE on the shared CQL session - it's easy to
create a separate session to use the "USE" on. With this fix, the test
no longer leaves the shared CQL session in a "USE" state.
2. Add a new xfailing test to reproduce the DESC KEYSPACES bug.
Refs #26334
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
Closesscylladb/scylladb#26345
The purpose of this two-patch series is to reproduce a previously unknown bug, Refs #26914.
Recently we saw a lot of patches that change how we create new schemas (keyspaces and tables), sometimes changing various long-time defaults. We started to worry that perhaps some of these defaults were applied only to CQL base tables and perhaps not to Alternator or to CQL's auxiliary tables (materialized views, secondary indexes, or CDC logs). For example, in Refs #26307 we wondered if perhaps the default "speculative_retry" option is different in Alternator than in CQL.
The first patch includes Alternator tests, and the second CQL tests. In both tests we discover that although recently (commit adf9c42, Refs #26610) we changed the default sstable compressor from LZ4Compressor to LZ4WithDictsCompressor, actually this change was **only** applied to CQL base tables. All Alternator tables and all CQL auxiliary tables (views, indexes, CDC) still use the old LZ4Compressor. This is issue #26914.
Closesscylladb/scylladb#26915
* github.com:scylladb/scylladb:
test/cqlpy: test compression setting for auxiliary table
test/alternator: tests for schema of Alternator table
This change adds support for secondary vector store clients, typically
located in different availability zones. Secondary clients serve as
fallback targets when all primary clients are unavailable.
New configuration option allows specifying secondary client addresses
and ports.
Fixes: VECTOR-187
Closesscylladb/scylladb#26484
When streaming SSTables by tablet range, the original implementation of tablet_sstable_streamer::stream may break out of the loop too early when encountering a non-overlaping SSTable. As a result, subsequent SSTables that should be classified as partially contained are skipped entirely.
Tablet range: [4, 5]
SSTable ranges:
[0,5]
[0, 3] <--- is considered exhausted, and causes skip to next tablet
[2, 5] <--- is missed for range [4, 5]
The loop uses if (!overlaps) break; semantics, which conflated “no overlap” with “done scanning.” This caused premature termination when an SSTable did not overlapped but the following one did.
Correct logic should be:
before(sst_last) → skip and continue.
after(sst_first) → break (no further SSTables can overlap).
Otherwise → `contains` to classify as full or partial.
Missing SSTables in streaming and potential data loss or incomplete streaming in repair/streaming operations.
1. Correct the loop termination logic that previously caused certain SSTables to be prematurely excluded, resulting in lost mutations. This change ensures all relevant SSTables are properly streamed and their mutations preserved.
2. Refactor the loop to use before() and after() checks explicitly, and only break when the SSTable is entirely after the tablet range
3. Add pytest to cover this case, full streaming flow by means of `restore`
4. Add boost tests to test the new refactored function
This data corruption fix should be ported back to 2024.2, 2025.1, 2025.2, 2025.3 and 2025.4
Fixes: https://github.com/scylladb/scylladb/issues/26979Closesscylladb/scylladb#26980
* github.com:scylladb/scylladb:
streaming: fix loop break condition in tablet_sstable_streamer::stream
streaming: add pytest case to reproduce mutation loss issue
vector_search: Fix error handling and status parsing
This change addresses two issues in the vector search client that caused
validator test failures: incorrect handling of 5xx server errors and
faulty status response parsing.
1. 5xx Error Handling:
Previously, a 5xx response (e.g., 503 Service Unavailable) from the
underlying vector store for an `/ann` search request was incorrectly
interpreted as a node failure. This would cause the node to be marked
as down, even for transient issues like an index scan being in progress.
This change ensures that 5xx errors are treated as transient search
failures, not node failures, preventing nodes from being incorrectly
marked as down.
2. Status Response Parsing:
The logic for parsing status responses from the vector store was
flawed. This has been corrected to ensure proper parsing.
Fixes: SCYLLADB-50
Backport to 2025.4 as this problem is present on this branch.
Closesscylladb/scylladb#27111
* github.com:scylladb/scylladb:
vector_search: Don't mark nodes as down on 5xx server errors
test: vector_search: Move unavailable_server to dedicated file
vector_search: Fix status response parsing
This patch adds a metric for pre-compression size of sstable files.
This patch adds a per-table metric
`scylla_column_family_total_disk_space_before_compression`,
which measures the hypothetical total size of sstables on disk,
if Data.db was replaced with an uncompressed equivalent.
As for the implementation:
Before the patch, tables and sstable sets are already tracking their total physical file size.
Whenever sstables are added or removed, the size delta is propagated from the sstable up through sstable sets into table_stats.
To implement the new metric, we turn the size delta that is getting passed around from a one-dimensional to a two-dimensional value, which includes both the physical and the pre-compression size.
New functionality, no backport needed.
Closesscylladb/scylladb#26996
* github.com:scylladb/scylladb:
replica/table: add a metric for hypothetical total file size without compression
replica/table: keep track of total pre-compression file size
For an `/ann` search request, a 5xx server response does not
indicate that the node is down. It can signify a transient state, such
as the index full scan being in progress.
Previously, treating a 503 error as a node fault would cause the node
to be incorrectly marked as down, for example, when a new index was
being created. This commit ensures that such errors are treated as
transient search failures, not node failures.
In pull request #26960 there was some discussion about what is the valid form of ExclusiveStartKey, and whether we need to allow some "non-standard" uses of it for scan over system tables (which aren't real Alternator tables and may have multiple key columns, something not possible in normal Altenrator tables). This made me realize our tests for what is allowed - and what is not allowed - in ExclusiveStartKey - are very sparse and don't cover all the cases that are possible in Scan and Query, in base tables and in GSIs.
So this small series attempts to increase the coverage of the tests for ExclusiveStartKey to make sure we are compatible with DynamoDB and also that we don't regress in #26960.
The new tests reproduce a previously unknown error-path issues, #26988, where in some cases DynamoDB considers ExclusiveStartKey to be invalid but Alternator erronously accepts. Fortunately, we didn't find any success-path (correctness) bugs.
Closesscylladb/scylladb#26994
* github.com:scylladb/scylladb:
test/alternator: tests for ExclusiveStartKey in GSI
test/alternator: more tests for ExclusiveStartKey in Scan
test/alternator: more tests for ExclusiveStartKey in Query
Increase the value of the max-networking-io-control-blocks option
for the cpp tests as it is too low and causes flakiness
as seen in vector_search.vector_store_client_test.vector_store_client_single_status_check_after_concurrent_failures:
```
seastar/src/core/reactor_backend.cc:342: void seastar::aio_general_context::queue(linux_abi::iocb *): Assertion `last < end` failed.
```
See also https://github.com/scylladb/seastar/issues/976Fixes#27056
Signed-off-by: Benny Halevy <bhalevy@scylladb.com>
Closesscylladb/scylladb#27117
Correct the loop termination logic that previously caused
certain SSTables to be prematurely excluded, resulting in
lost mutations. This change ensures all relevant SSTables
are properly streamed and their mutations preserved.
More efficient than 100 pings.
There was one ping in test which was done "so this shard notices the
clock advance". It's not necessary, since obsering completed SMP
call implies that local shard sees the clock advancement done within in.
Primary issue with the old method is that each update is a separate
cross-shard call, and all later updated queue behind it. If one of the
shards has high latency for such calls, the queue may accumulate and
system will appear unresponsive for mapping changes on non-zero shards.
This happened in the field when one of the shards was overloaded with
sstables and compaction work, which caused frequent stalls which
delayed polling for ~100ms. A queue of 3k address updates
accumulated. This made bootstrap impossible, since nodes couldn't
learn about the IP mapping for the bootstrapping node and streaming
failed.
To protect against that, use a more efficient method of replication
which requires a single cross-shard call to replicate all prior
updates.
It is also more reliable, if replication fails transiently for some
reason, we don't give up and fail all later updates.
Fixes#26865Fixes#26835
Key goals:
- efficient (batching updates)
- reliable (no lost updates)
Will be used in data structures maintained on one designed owning
shard and replicated to other shards.
Somehow, the line of code responsible for freeing flushed nodes
in `trie_writer` is missing from the implementation.
This effectively means that `trie_writer` keeps the whole index in
memory until the index writer is closed, which for many dataset
is a guaranteed OOM.
Fix that, and add some test that catches this.
Fixesscylladb/scylladb#27082Closesscylladb/scylladb#27083
The response was incorrectly parsed as a plain string and compared
directly with C++ string. However, the body contains a JSON string,
which includes escaped quotes that caused comparison failures.
In the previous patch we noticed that although recently (commit adf9c42,
Refs #26610) we changed the default sstable compressor from LZ4Compressor
to LZ4WithDictsCompressor, this change was only applied to CQL, not to
Alternator.
In this patch we add tests that demonstrate that it's even worse - the
new compression only applies to CQL's *base* table - all the "auxiliary"
tables -
* Materialized views
* Secondary index's materialized views
* CDC log tables
all still have the old LZ4Compressor, different from the base table's
default compressor.
The new test fails on Scylla, reproducing #26914, and passes on
Cassandra (on Cassandra, we only compare the materialized view table,
because SI and CDC is implemented differently).
Signed-off-by: Nadav Har'El <nyh@scylladb.com>
