This commit fixes a number of issues with data replication:
- Local and global checkpoints are not updated after the new operations have been fsynced, but
might capture a state before the fsync. The reason why this probably went undetected for so
long is that AsyncIOProcessor is synchronous if you index one item at a time, and hence working
as intended unless you have a high enough level of concurrent indexing. As we rely in other
places on the assumption that we have an up-to-date local checkpoint in case of synchronous
translog durability, there's a risk for the local and global checkpoints not to be up-to-date after
replication completes, and that this won't be corrected by the periodic global checkpoint sync.
- AsyncIOProcessor also has another "bad" side effect here: if you index one bulk at a time, the
bulk is always first fsynced on the primary before being sent to the replica. Further, if one thread
is tasked by AsyncIOProcessor to drain the processing queue and fsync, other threads can
easily pile more bulk requests on top of that thread. Things are not very fair here, and the thread
might continue doing a lot more fsyncs before returning (as the other threads pile more and
more on top), which blocks it from returning as a replication request (e.g. if this thread is on the
primary, it blocks the replication requests to the replicas from going out, and delaying
checkpoint advancement).
This commit fixes all these issues, and also simplifies the code that coordinates all the after
write actions.
Some unit test checking locale sensitive functionality require the
-Djava.locale.providers=SPI,COMPAT flag to be set. When running tests though
gradle we pass this already to the BuildPlugin, but running from the IDE this
might need to be set manually. Adding a note explaining this to the
CONTRIBUTING.md doc and leaving a note in the test comment of
SearchQueryIT.testRangeQueryWithLocaleMapping which is a test we know
that suffers from this issue.
Reindex sort never gave a guarantee about the order of documents being
indexed into the destination, though it could give a sense of locality
of source data.
It prevents us from doing resilient reindex and other optimizations and
it has therefore been deprecated.
Related to #47567
Some queries return bulk scorers that can be significantly faster than
iterating naively over the scorer. By giving script_score a BulkScorer
that would delegate to the wrapped query, we could make it faster in some cases.
Closes#40837
This rewrites long sort as a `DistanceFeatureQuery`, which can
efficiently skip non-competitive blocks and segments of documents.
Depending on the dataset, the speedups can be 2 - 10 times.
The optimization can be disabled with setting the system property
`es.search.rewrite_sort` to `false`.
Optimization is skipped when an index has 50% or more data with
the same value.
Optimization is done through:
1. Rewriting sort as `DistanceFeatureQuery` which can
efficiently skip non-competitive blocks and segments of documents.
2. Sorting segments according to the primary numeric sort field(#44021)
This allows to skip non-competitive segments.
3. Using collector manager.
When we optimize sort, we sort segments by their min/max value.
As a collector expects to have segments in order,
we can not use a single collector for sorted segments.
We use collectorManager, where for every segment a dedicated collector
will be created.
4. Using Lucene's shared TopFieldCollector manager
This collector manager is able to exchange minimum competitive
score between collectors, which allows us to efficiently skip
the whole segments that don't contain competitive scores.
5. When index is force merged to a single segment, #48533 interleaving
old and new segments allows for this optimization as well,
as blocks with non-competitive docs can be skipped.
Backport for #48804
Co-authored-by: Jim Ferenczi <jim.ferenczi@elastic.co>
Reindex sort never gave a guarantee about the order of documents being
indexed into the destination, though it could give a sense of locality
of source data.
It prevents us from doing resilient reindex and other optimizations and
it has therefore been deprecated.
Related to #47567
This stems from a time where index requests were directly forwarded to
TransportReplicationAction. Nowadays they are wrapped in a BulkShardRequest, and this logic is
obsolete.
In contrast to prior PR (#49647), this PR also fixes (see b3697cc) a situation where the previous
index expression logic had an interesting side effect. For bulk requests (which had resolveIndex
= false), the reroute phase was waiting for the index to appear in case where it was not present,
and for all other replication requests (resolveIndex = true) it would right away throw an
IndexNotFoundException while resolving the name and exit. With #49647, every replication
request was now waiting for the index to appear, which was problematic when the given index
had just been deleted (e.g. deleting a follower index while it's still receiving requests from the
leader, where these requests would now wait up to a minute for the index to appear). This PR
now adds b3697cc on top of that prior PR to make sure to reestablish some of the prior behavior
where the reroute phase waits for the bulk request for the index to appear. That logic was in
place to ensure that when an index was created and not all nodes had learned about it yet, that
the bulk would not fail somewhere in the reroute phase. This is now only restricted to the
situation where the current node has an older cluster state than the one that coordinated the
bulk request (which checks that the index is present). This also means that when an index is
deleted, we will no longer unnecessarily wait up to the timeout for the index o appear, and
instead fail the request.
Closes#20279
* Make BlobStoreRepository Aware of ClusterState (#49639)
This is a preliminary to #49060.
It does not introduce any substantial behavior change to how the blob store repository
operates. What it does is to add all the infrastructure changes around passing the cluster service to the blob store, associated test changes and a best effort approach to tracking the latest repository generation on all nodes from cluster state updates. This brings a slight improvement to the consistency
by which non-master nodes (or master directly after a failover) will be able to determine the latest repository generation. It does not however do any tricky checks for the situation after a repository operation
(create, delete or cleanup) that could theoretically be used to get even greater accuracy to keep this change simple.
This change does not in any way alter the behavior of the blobstore repository other than adding a better "guess" for the value of the latest repo generation and is mainly intended to isolate the actual logical change to how the
repository operates in #49060
This commit adds a function in NodeClient that allows to track the progress
of a search request locally. Progress is tracked through a SearchProgressListener
that exposes query and fetch responses as well as partial and final reduces.
This new method can be used by modules/plugins inside a node in order to track the
progress of a local search request.
Relates #49091
This stems from a time where index requests were directly forwarded to
TransportReplicationAction. Nowadays they are wrapped in a BulkShardRequest, and this logic is
obsolete.
Closes#20279
This change adds a dynamic cluster setting named `indices.id_field_data.enabled`.
When set to `false` any attempt to load the fielddata for the `_id` field will fail
with an exception. The default value in this change is set to `false` in order to prevent
fielddata usage on this field for future versions but it will be set to `true` when backporting
to 7x. When the setting is set to true (manually or by default in 7x) the loading will also issue
a deprecation warning since we want to disallow fielddata entirely when https://github.com/elastic/elasticsearch/issues/26472
is implemented.
Closes#43599
JavaDateFormatter should keep the pattern with the prefixed 8 as it will be used for serialisation. The stripped pattern should be used for the enclosed formatters.
closes#48698
Fixes a bug where a scripted upsert that causes a dynamic mapping update is retried (because
mapping update is still in-flight), and the request is mutated multiple times.
Closes#48670
Backport of #49076
In case an exception occurs inside a pipeline processor,
the pipeline stack is kept around as header in the exception.
Then in the on_failure processor the id of the pipeline the
exception occurred is made accessible via the `on_failure_pipeline`
ingest metadata.
Closes#44920
Preliminary to shorten the diff of #49060. In #49060
we execute cluster state updates during the writing of a new
index gen and thus it must be an async API.
All the implementations of `EsBlobStoreTestCase` use the exact same
bootstrap code that is also used by their implementation of
`EsBlobStoreContainerTestCase`.
This means all tests might as well live under `EsBlobStoreContainerTestCase`
saving a lot of code duplication. Also, there was no HDFS implementation for
`EsBlobStoreTestCase` which is now automatically resolved by moving the tests over
since there is a HDFS implementation for the container tests.
The annotated text mapper has a field type that currently extends StringFieldType,
which means that all the positional-related query factory methods need to be copied
over from TextFieldType. In addition, MappedFieldType.intervals() hasn't been
overridden, so you can't use intervals queries with annotated text - a major drawback,
since one of the purposes of annotated text is to be able to run positional queries against
annotations.
This commit changes the annotated text field type to extend TextFieldType instead,
adding tests to ensure that position queries work correctly.
Closes#49289
Same as #49518 pretty much but for GCS.
Fixing a few more spots where input stream can get closed
without being fully drained and adding assertions to make sure
it's always drained.
Moved the no-close stream wrapper to production code utilities since
there's a number of spots in production code where it's also useful
(will reuse it there in a follow-up).
Backport of #49552.
Closes#49428. The code that works out an HTTP code for an exception didn't
consider the JsonParseException case, meant that an invalid JSON request could
result in a 500 Internal Server Error. Now it returns 400 Bad Request.
This commit back ports three commits related to enabling the simple
connection strategy.
Allow simple connection strategy to be configured (#49066)
Currently the simple connection strategy only exists in the code. It
cannot be configured. This commit moves in the direction of allowing it
to be configured. It introduces settings for the addresses and socket
count. Additionally it introduces new settings for the sniff strategy
so that the more generic number of connections and seed node settings
can be deprecated.
The simple settings are not yet registered as the registration is
dependent on follow-up work to validate the settings.
Ensure at least 1 seed configured in remote test (#49389)
This fixes#49384. Currently when we select a random subset of seed
nodes from a list, it is possible for 0 seeds to be selected. This test
depends on at least 1 seed being selected.
Add the simple strategy to cluster settings (#49414)
This is related to #49067. This commit adds the simple connection
strategy settings and strategy mode setting to the cluster settings
registry. With these changes, the simple connection mode can be used.
Additionally, it adds validation to ensure that settings cannot be
misconfigured.
The new CompensatedSum is a nice DRY refactor, but had the unanticipated
side effect of creating a lot of object allocation in the aggregation hot collection
loop: one object per visited document, per aggregator. In some places it
created two per-doc-per-agg (weighted avg, geo centroids, etc) since there
were multiple compensations being maintained.
This PR moves the object creation out of the hot loop so that it is now
created once per segment, and resets the internal state each time through
the loop
A few enhancements to `SnapshotResiliencyTests`:
1. Test running requests from random nodes in more spots to enhance coverage (this is particularly motivated by #49060 where the additional number of cluster state updates makes it more interesting to fully cover all kinds of network failures)
2. Fix issue with restarting only master node in one test (doing so breaks the test at an incredibly low frequency, that becomes not so low in #49060 with the additional cluster state updates between request and response)
3. Improved cluster formation checks (now properly checks the term as well when forming cluster) + makes sure all nodes are connected to all other nodes (previously the data nodes would at times not be connected to other data nodes, which was shaken out now by adding the `client()` method
4. Make sure the cluster left behind by the test makes sense by running the repo cleanup action on it (this also increases coverage of the repository cleanup action obviously and adds the basis of making it part of more resiliency tests)
The NodeTests class contains tests that check behavior when shutting
down a node. This involves starting a node, performing some operation,
stopping the node, and then awaiting the close of the node. Part of
closing a node is the termination of the node's ThreadPool. ThreadPool
termination semantics can be deceiving. The ThreadPool#terminate method
takes a timeout value and the first oddity is that the terminate method
can take two times the timeout value before returning. Internally this
method acts on the ExecutorService instances that are held by the
ThreadPool. First, an orderly shutdown is attempted and pending tasks
are allowed to execute while waiting for the timeout value. If any of
the ExecutorService instances have not terminated, a call is made to
attempt to stop all active tasks (usually using interrupts) and then
waits for up to the timeout value a second time for the termination of
the ExecutorService instances. This means that if use a large value
when waiting for a node to close, we may not attempt to interrupt any
threads that are in a blocking call before the test times out.
In order to avoid causing these tests to time out, this change reduces
the timeout passed to Node#awaitClose to 10 seconds from 1 day. This
will allow blocked threads to be interrupted before the test suite
fails due to the timeout.
Closes#44256Closes#42350Closes#44435
This commit enhances the required pipeline functionality by changing it
so that default/request pipelines can also be executed, but the required
pipeline is always executed last. This gives users the flexibility to
execute their own indexing pipelines, but also ensure that any required
pipelines are also executed. Since such pipelines are executed last, we
change the name of required pipelines to final pipelines.
Just realized we were missing some annotations here which was somewhat
confusing since other methods/parameters have the `Nullable` annotation
wherever a `null` can be passed.
Backport of #47208.
Closes#46900. When running ES with `--quiet`, if ES then exits abnormally, a
user has to go hunting in the logs for the error. Instead, never close
System.err, and print more information to it if ES encounters a fatal error
e.g. config validation, or some fatal runtime exception. This is useful when
running under e.g. systemd, since the error will go into the journal.
Note that stderr is still closed in daemon (`-d`) mode.
This change automatically pre-sort search shards on search requests that use a primary sort based on the value of a field. When possible, the can_match phase will extract the min/max (depending on the provided sort order) values of each shard and use it to pre-sort the shards prior to running the subsequent phases. This feature can be useful to ensure that shards that contain recent data are executed first so that intermediate merge have more chance to contain contiguous data (think of date_histogram for instance) but it could also be used in a follow up to early terminate sorted top-hits queries that don't require the total hit count. The latter could significantly speed up the retrieval of the most/least
recent documents from time-based indices.
Relates #49091
Currently the condtion that is supposed to test creation of test instances with
multiple indices is never true because it compares Strings with an enum. This
changes it so the condition uses the enum constants instead.
Lucene 8.3 included a root fix for #43976, which was temporarily fixed in elasticsearch
by #44340. Since we have upgraded to 8.3 we no longer need this workaround. This
commit fixes the test that was added to check the workaround, and instead checks that
fields with index_phrases enabled correctly build queries when used with multi-term
synonyms.
Closes#47777
When opening a translog file, we check whether the UUID matches what we expect (the UUID
from the latest commit). The UUID check can in certain cases fail when the translog is
corrupted. This commit changes the ordering of the checks so that corruption is detected first.
If the translog on a replica is corrupt, we should not perform an
operation-based recovery or utilize sync_id as we won't be able to open
an engine in the next step. This change adds an extra validation that
ensures translog is okay when preparing a peer recovery request.
This change disables the query and request cache when
profile is set to true in the request. This means that profiled queries
will not check caches to execute the query and the result will never be
added in the cache either.
Closes#33298
Yannick Welsch
Mayya Sharipova
jimczi
Armin Braun
Ignacio Vera
Christoph Büscher
Henning Andersen
Henning Andersen
Adrien Grand
Christos Soulios
Martijn van Groningen
Przemyslaw Gomulka
Alan Woodward
Rory Hunter
Tim Brooks
Zachary Tong
Jared Tan
Jason Tedor
Jay Modi
Igor Motov
Nhat Nguyen
jaymode