At the shard level we use an operation permit to coordinate between regular shard operations and special operations that need exclusive access. In ES versions < 6, the operation requiring exclusive access was invoked during primary relocation, but ES versions >= 6 this exclusive access is also used when a replica learns about a new primary or when a replica is promoted to primary.
These special operations requiring exclusive access delay regular operations from running, by adding them to a queue, and after finishing the exclusive access, release these operations which then need to be put back on the original thread-pool they were running on. In the presence of thread pool rejections, the current implementation had two issues:
- it would not properly release the operation permit when hitting a rejection (i.e. when calling ThreadedActionListener.onResponse from IndexShardOperationPermits.acquire).
- it would not invoke the onFailure method of the action listener when the shard was closed, and just log a warning instead (see ThreadedActionListener.onFailure), which would ultimately lead to the replication task never being cleaned up (see #25863).
This commit fixes both issues by introducing a custom threaded action listener that is permit-aware and properly deals with rejections.
Closes#25863
It fixes random score generation to ensure that you will not always get the
same scores on a read-only index by integrating the seed into the score
computation when using doc ids. It also removes `ctx.docBase` from the formula
since it might change over time if deletes are compacted while scores are
supposed to be cacheable per segment.
Extracts ranges from range queries on byte, short, integer, long, half_float, scaled_float, float, double, date and ip fields.
byte, short, integer and date ranges are normalized to Lucene's LongRange.
half_float and float are normalized to Lucene's DoubleRange.
When extracting range queries, the QueryAnalyzer computes the width of the range. This width is used to determine
what range should be preferred in a conjunction query. The QueryAnalyzer prefers the smaller ranges, because these
ranges tend to match with less documents.
Closes#21040
Today we expose `IndexFieldDataService` outside of IndexService to do maintenance
or lookup field data in different ways. Yet, we have a streamlined way to access IndexFieldData
via `QueryShardContext` that should encapsulate all access to it. This also ensures that we control all other functionality like cache clearing etc.
This change also removes the `recycler` option from `ClearIndicesCacheRequest` this option is a no-op and should have been removed long ago.
Currently, NioTransport does start normal socket selectors and the
client when the network server setting is set to false. This commit
makes it so that the client will be started even when the network server
is not enabled.
Additionally, it randomly introduces the NioTransport as an option for
the MockTransportClient throughout tests.
This predicate is used to deal with the intricacies of detecting when a master is reelected/nodes rejoins an existing master. The current implementation is based on nodeIds, which is fine if the master really change. If the nodeId is equal the code falls back to detecting an increment in the cluster state version which happens when a node is re-elected or when the node rejoins. Sadly this doesn't cover the case where the same node is elected after a full restart of all master nodes. In that case we recover the cluster state from disk but the version is reset back to 0. To fix this, the check should be done based on ephemeral IDs which are reset on restart.
Fixes#25471
These two methods do do the same thing. The subtle difference between the two is that the former prints out pretty printed content by default while the latter doesn't. There are way more usages of the latter throughout the codebase hence I kept that variant although I do think that it would be much better to print out prettified content by default from a `toString`. That breaks quite some tests so I didn't make that change yet.
Also XContentHelper#toString was outdated as it didn't check the ToXContent#isFragment method to decide whether a new anonymous object has to be created or not. It would simply fail with any ToXContentObject.
The test only waited for one op to be stuck. In rare occasions the other ops were still in flight when recovery captured a translog snapshot throwing doc count off.
The configuration removed from the runtime configuration did not
properly remove the deps jar from gradle versions > 3.3. The rest client
now removes both the 3.3 and 3.3+ configurations so this works on both
versions of gradle.
Closes#25884
Relates #25208
Today when we aggregate on the `_index` field the cross cluster search
alias is not taken into account. Neither is it respected when we search
on the field. This change adds support for cluster alias when the cluster
alias is present on the `_index` field.
Closes#25606
This changes makes it so you can index a value like "1.0" or "1.1" into whole
number field types like byte and integer. Without this change then the above
values would have resulted in an error, even with coerce set to true.
Closes#25819
Currently we have an option to interrupt the selector thread on close.
This option is not needed as we do not call this method and we should
not be blocking on the network thread. Instead we only need to ever call
wakeup() on the raw selector.
We cannot guarantee that the result of computations will be in the float range,
since it depends on the data and how scores are computed. We already use doubles
as intermediate representations and cast to a float as a final step, which is
the right thing to do. Small doubles will just be rounded to zero, there is not
much we can or should do about it.
Closes#25330
Stored fields were still being accessed for nested inner hits even if the _source was not requested.
This was done to figure out the id of the root document. However this is already known higher up the stack.
So instead this change adds the id to the nested search context, so that it is no longer required to be fetched via the stored fields.
In case the _source is large and no source is requested then hot threads like these ones would still appear:
```
100.3% (501.3ms out of 500ms) cpu usage by thread 'elasticsearch[AfXKKfq][search][T#6]'
2/10 snapshots sharing following 22 elements
org.apache.lucene.store.DataInput.skipBytes(DataInput.java:352)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader.skipField(CompressingStoredFieldsReader.java:246)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader.visitDocument(CompressingStoredFieldsReader.java:601)
org.apache.lucene.index.CodecReader.document(CodecReader.java:88)
org.apache.lucene.index.FilterLeafReader.document(FilterLeafReader.java:411)
org.elasticsearch.search.fetch.FetchPhase.loadStoredFields(FetchPhase.java:347)
org.elasticsearch.search.fetch.FetchPhase.createNestedSearchHit(FetchPhase.java:219)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:150)
org.elasticsearch.search.fetch.subphase.InnerHitsFetchSubPhase.hitsExecute(InnerHitsFetchSubPhase.java:73)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:166)
org.elasticsearch.search.fetch.subphase.InnerHitsFetchSubPhase.hitsExecute(InnerHitsFetchSubPhase.java:73)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:166)
org.elasticsearch.search.SearchService.executeFetchPhase(SearchService.java:422)
```
and:
```
8/10 snapshots sharing following 27 elements
org.apache.lucene.codecs.compressing.LZ4.decompress(LZ4.java:135)
org.apache.lucene.codecs.compressing.CompressionMode$4.decompress(CompressionMode.java:138)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader$BlockState$1.fillBuffer(CompressingStoredFieldsReader.java:531)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader$BlockState$1.readBytes(CompressingStoredFieldsReader.java:550)
org.apache.lucene.store.DataInput.readBytes(DataInput.java:87)
org.apache.lucene.store.DataInput.skipBytes(DataInput.java:350)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader.skipField(CompressingStoredFieldsReader.java:246)
org.apache.lucene.codecs.compressing.CompressingStoredFieldsReader.visitDocument(CompressingStoredFieldsReader.java:601)
org.apache.lucene.index.CodecReader.document(CodecReader.java:88)
org.apache.lucene.index.FilterLeafReader.document(FilterLeafReader.java:411)
org.elasticsearch.search.fetch.FetchPhase.loadStoredFields(FetchPhase.java:347)
org.elasticsearch.search.fetch.FetchPhase.createNestedSearchHit(FetchPhase.java:219)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:150)
org.elasticsearch.search.fetch.subphase.InnerHitsFetchSubPhase.hitsExecute(InnerHitsFetchSubPhase.java:73)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:166)
org.elasticsearch.search.fetch.subphase.InnerHitsFetchSubPhase.hitsExecute(InnerHitsFetchSubPhase.java:73)
org.elasticsearch.search.fetch.FetchPhase.execute(FetchPhase.java:166)
org.elasticsearch.search.SearchService.executeFetchPhase(SearchService.java:422)
```
Currently Engine.close can return immediately if the engine is already at the process of shutting down (due to a concurrent close call or an engine failure). This is a shame because some of our testing infra wants to do things like checking the index. This commit changes the logic to make sure that all calls to close wait until resources are freed. Failing the engine is still non blocking.
Fixes#25817
This commit removes all external dependencies from the rest client jar
and shades them in an 'org.elasticsearch.client' package within the jar
using shadowJar gradle plugin. All projects that depended on the
existing jar have been converted to using the 'org.elasticsearch.client'
package prefixes to interact with the rest client.
Closes#25208
This change disables the graph analysis on default `shingle` filter.
The pre-configured shingle filter produces shingles of different size.
Graph analysis on such token stream is useless and dangerous as it may create too many paths.
Fixes#25555
The context suggester extracts the context field values from the document but it does not filter doc values field coming from Keyword field.
This change filters doc values field when building the context values.
Fixes#25404
This change handles the case where a SpanNearQueryBuilder tries to create a query with a single clause.
This is not allowed in the SpanNearQuery so instead of throwing an exception when the weight is built, this change builds and returns
the singleton inner clause on toQuery.
Fixes#25630
The default _parent field tries to load global ordinals because it is created with eager_global_ordinals=true.
This leads to an IllegalStateException because this field does not have doc_values.
This change explicitely sets eager_global_ordinals to false in order to avoid the ISE on startup.
Fixes#25849
Currently we are failing to close socket channels when the initial bind
or connect operation fails. This leaves the file descriptor hanging
around. This closes the channel when an exception occurs during bind or
connect.
Currently an NioChannel is created and it is UNREGISTERED. At some point
it is registered with a selector. From that point on, the channel can
only be closed by the selector. The fact that a channel might not be
associated with a selector has significant implications for concurrency
and the channel shutdown process. The only thing that is simplified by
allowing channels to be in a state independent of a selector is some
testing scenarios.
This PR modifies channels so that they are given a selector at creation
time and are always associated with that selector. Only that selector
can close that channel. This simplifies the channel lifecycle and
closing intricacies.
Since the setup attempts to create an index with two types, and the setup runs before any test,
this will fail on versions 6.0+ before it has a chance to check the skip in each individual
test. Moving to the setup resolves this issue.
When a replica processes out of order operations, it can drop some due to version comparisons. In the past that would have resulted in a VersionConflictException being thrown and the operation was totally ignored. With the seq# push, we started storing these operations in the translog (but not indexing them into lucene) in order to have complete op histories to facilitate ops based recoveries. This in turn had the undesired effect that deleted docs may be resurrected during recovery in some extreme edge situation (see a complete explanation below). This PR contains a simple fix, which is also an optimization for the recovery process, incoming operation that have a seq# lower than the current local checkpoint (i.e., have already been processed) should not be indexed into lucene. Note that sometimes we can also skip storing them in the translog, but this is not required for the fix and is more complicated.
This is the equivalent of #25592
## More details on resurrected ops
Consider two operations:
- Index d1, seq no 1
- Delete d1, seq no 3
On a replica they come out of order:
- Translog gen 1 contains:
- delete (seqNo 3)
- Translog gen 2 contains:
- index (seqNo 1) (wasn't indexed into lucene, but put into the translog)
- another operation (seqNo 10)
- Translog gen 3
- another op (seqNo 9)
- Engine commits with:
- local checkpoint 9
- refers to gen 2
If this replica becomes a primary:
- Local recovery will replay translog gen 2 and up, causing index #1 to be re-index.
- Even if recovery will start at gen 3, the translog retention policy will cause file based recovery to replay the entire translog. If it happens to start at gen 2 (but not 1), we will run into the same problem.
#### Some context - out of order delivery involving deletes:
On normal operations, this relies on the gc_deletes setting. We assume that the setting represents an upper bound on the time between the index and the delete operation. The index operation will be detected as stale based on the tombstone map in the LiveVersionMap.
Recovery presents a challenge as it can replay an old index operation that was in the translog and override a delete operation that was done when the engine was opened (and is not part of the replayed snapshot). To deal with this situation, we disable GC deletes (i.e. retain all deletes) for the duration of recoveries. This means that the delete operation will be remembered and the index operation ignored.
Both of the above scenarios (local recover + peer recovery) create a situation where the delete operation is never replayed. It this "lost" as lucene doesn't remember it happened and our LiveVersionMap is populated with it.
#### Solution:
Note that both local and peer recovery represent a scenario where we replay translog ops on top of an existing lucene index, potentially with ongoing indexing. Therefore we can treat them the same.
The local checkpoint in Lucene represent a marker indicating that all operations below it were performed on the index. This is the only form of "memory" that we have that relates to deletes. If we can achieve the following:
1) All ops below the local checkpoint are not indexed to lucene.
2) All ops above the local checkpoint are
It will mean that all variants are covered: (i# == index op seq#, d# == delete op seq#, lc == local checkpoint in commit)
1) i# < d# <= lc - document is already deleted in lucene and stays that way.
2) i# <= lc < d# - delete is replayed on index - document is deleted
3) lc < i# < d# - index is replayed and then delete - document is deleted.
More formally - we want to make sure that for all ops that performed on the primary o1 and o2, if o2 is processed on a shard before o1, o1 will be dropped. We have the following scenarios
1) If both o1 or o2 are not included in the replayed snapshot and are above it (i.e., have a higher seq#), they fall under the gc deletes assumption.
2) If both o1 is part of the replayed snapshot but o2 is above it:
- if o2 arrives first, o1 must arrive due to the recovery and potentially via replication as well. since gc deletes is disabled we are guaranteed to know of o2's existence.
3) If both o2 and o1 are part of the replayed snapshot:
- we fall under the same scenarios as #2 - disabling GC deletes ensures we know of o2 if it arrives first.
4) If o1 falls before the snapshot and o2 is either part of the snapshot or higher:
- Since the snapshot is guaranteed to contain all ops that are not part of lucene and are above the lc in the commit used, this means that o1 is part of lucene and o1 < local checkpoint. This means it won't be processed and we're not in the scenario we're discussing.
5) If o2 falls before the snapshot but o1 is part of it:
- by the same reasoning above, o2 is < local checkpoint. Since o1 < o2, we also get o1 < local checkpoint and this will be dropped.
#### Implementation:
For local recovery, we can filter the ops we read of the translog and avoid replaying them. For peer recovery this is tricky as we do want to send the operations in order to have some history on the target shard. Filtering operations on the engine level (i.e., not indexing to lucene if op seq# <= lc) would work for both.
This commit changes the way we handle field expansion in `match`, `multi_match` and `query_string` query.
The main changes are:
- For exact field name, the new behavior is to rewrite to a matchnodocs query when the field name is not found in the mapping.
- For partial field names (with `*` suffix), the expansion is done only on `keyword`, `text`, `date`, `ip` and `number` field types. Other field types are simply ignored.
- For all fields (`*`), the expansion is done on accepted field types only (see above) and metadata fields are also filtered.
- The `*` notation can also be used to set `default_field` option on`query_string` query. This should replace the needs for the extra option `use_all_fields` which is deprecated in this change.
This commit also rewrites simple `*` query to matchalldocs query when all fields are requested (Fixes#25556).
The same change should be done on `simple_query_string` for completeness.
`use_all_fields` option in `query_string` is also deprecated in this change, `default_field` should be set to `*` instead.
Relates #25551
Removes the primary term from the replication request and pushes it into the transport envelope. This makes it possible to remove the term from the ReplicationOperation universe. The primary term that is to be used for a replication operation is now determined in the reroute phase when the node decides to execute a primary action (and validated once the primary action gets to execute). This makes it possible to validate that the primary action was sent to the correct primary shard instance that it was meant to be sent to (currently we only validate primary actions using the allocation id, which can be reused for failed and reallocated primaries).
If a primary shard is relocated, and then subsequently closed, there is a short window where ReplicationOperation could access the
closed shard (engine is not shut down yet) and, because it does not know that the shard was relocated, try to update the local
checkpoint, tripping an assertion in GlobalCheckPointTracker that a local checkpoint cannot be updated if it's not in primary mode.
Yannick Welsch
javanna
Adrien Grand
Martijn van Groningen
Simon Willnauer
Tim Brooks
Boaz Leskes
Luca Cavanna
Tanguy Leroux
Michael Basnight
Scott Somerville
Lee Hinman
Clinton Gormley
Jim Ferenczi
Zachary Tong
Colin Goodheart-Smithe