For too long we have been groping around in the dark when faced with GC
issues because we rarely have GC logs at our disposal. This commit
enables GC logging by default out of the box.
Relates #27610
Any CLI commands that depend on core Elasticsearch might touch classes
(directly or indirectly) that depends on logging. If they do this and
logging is not configured, Log4j will dump status error messages to the
console. As such, we need to ensure that any such CLI command configures
logging (with a trivial configuration that dumps log messages to the
console). Previously we did this in the base CLI command but with the
refactoring of this class out of core Elasticsearch, we no longer
configure logging there (since we did not want this class to depend on
settings and logging). However, this meant for some CLI commands (like
the plugin CLI) we were no longer configuring logging. This commit adds
base classes between the low-level command and multi-command classes
that ensure that logging is configured. Any CLI command that depends on
core Elasticsearch should use this infrastructure to ensure logging is
configured. There is one exception to this: Elasticsearch itself because
it takes reponsibility into its own hands for configuring logging from
Elasticsearch settings and log4j2.properties. We preserve this special
status.
Relates #27523
Today we require users to prepare their indices for split operations.
Yet, we can do this automatically when an index is created which would
make the split feature a much more appealing option since it doesn't have
any 3rd party prerequisites anymore.
This change automatically sets the number of routinng shards such that
an index is guaranteed to be able to split once into twice as many shards.
The number of routing shards is scaled towards the default shard limit per index
such that indices with a smaller amount of shards can be split more often than
larger ones. For instance an index with 1 or 2 shards can be split 10x
(until it approaches 1024 shards) while an index created with 128 shards can only
be split 3x by a factor of 2. Please note this is just a default value and users
can still prepare their indices with `index.number_of_routing_shards` for custom
splitting.
NOTE: this change has an impact on the document distribution since we are changing
the hash space. Documents are still uniformly distributed across all shards but since
we are artificually changing the number of buckets in the consistent hashign space
document might be hashed into different shards compared to previous versions.
This is a 7.0 only change.
Today Cross Cluster Search requires at least one node in each remote cluster to be up once the cross cluster search is run. Otherwise the whole search request fails despite some of the data (either local and/or remote) is available. This happens when performing the _search/shards calls to find out which remote shards the query has to be executed on. This scenario is different from shard failures that may happen later on when the query is actually executed, in case e.g. remote shards are missing, which is not going to fail the whole request but rather yield partial results, and the _shards section in the response will indicate that.
This commit introduces a boolean setting per cluster called search.remote.$cluster_alias.skip_if_disconnected, set to false by default, which allows to skip certain clusters if they are down when trying to reach them through a cross cluster search requests. By default all clusters are mandatory.
Scroll requests support such setting too when they are first initiated (first search request with scroll parameter), but subsequent scroll rounds (_search/scroll endpoint) will fail if some of the remote clusters went down meanwhile.
The search API response contains now a new _clusters section, similar to the _shards section, that gets returned whenever one or more clusters were disconnected and got skipped:
"_clusters" : {
"total" : 3,
"successful" : 2,
"skipped" : 1
}
Such section won't be part of the response if no clusters have been skipped.
The per cluster skip_unavailable setting value has also been added to the output of the remote/info API.
Projects the depend on the CLI currently depend on core. This should not
always be the case. The EnvironmentAwareCommand will remain in :core,
but the rest of the CLI components have been moved into their own
subproject of :core, :core:cli.
Currently, we are using a plain TransportRequestHandler to post snapshot
status messages to the master. However, it doesn't have a robust retry
mechanism as TransportMasterNodeAction. This change migrates from
TransportRequestHandler to TransportMasterNodeAction for the new
versions and keeps the current implementation for the old versions.
Closes#27151
When the vagrant box is very very slow, the elasticsearch service can
take more than 60 sec to start. This commit changes the timeout to 120.
closes#27372
* REST: Rename ingest.processor.grok to ingest.processor_grok
* REST: Rename remote.info to cluster.remote_info
* REST: Fixed bad YAML comments
* REST: Force dummy scripts to be strings, not numbers
* REST: Fix bad YAML in search/110_field_collapsing.yml
* REST: Adjust percentile tests to work with Perl number handling
extract all clauses from a conjunction query.
When clauses from a conjunction are extracted the number of clauses is
also stored in an internal doc values field (minimum_should_match field).
This field is used by the CoveringQuery and allows the percolator to
reduce the number of false positives when selecting candidate matches and
in certain cases be absolutely sure that a conjunction candidate match
will match and then skip MemoryIndex validation. This can greatly improve
performance.
Before this change only a single clause was extracted from a conjunction
query. The percolator tried to extract the clauses that was rarest in order
(based on term length) to attempt less candidate queries to be selected
in the first place. However this still method there is still a very high
chance that candidate query matches are false positives.
This change also removes the influencing query extraction added via #26081
as this is no longer needed because now all conjunction clauses are extracted.
https://www.elastic.co/guide/en/elasticsearch/reference/6.x/percolator.html#_influencing_query_extractionCloses#26307
If an out of memory error is thrown while merging, today we quietly
rewrap it into a merge exception and the out of memory error is
lost. Instead, we need to rethrow out of memory errors, and in fact any
fatal error here, and let those go uncaught so that the node is torn
down. This commit causes this to be the case.
Relates #27265
The warnings headers have a fairly limited set of valid characters
(cf. quoted-text in RFC 7230). While we have assertions that we adhere
to this set of valid characters ensuring that our warning messages do
not violate the specificaion, we were neglecting the possibility that
arbitrary user input would trickle into these warning headers. Thus,
missing here was tests for these situations and encoding of characters
that appear outside the set of valid characters. This commit addresses
this by encoding any characters in a deprecation message that are not
from the set of valid characters.
Relates #27269
Only tests should use the single argument Environment constructor. To
enforce this the single arg Environment constructor has been replaced with
a test framework factory method.
Production code (beyond initial Bootstrap) should always use the same
Environment object that Node.getEnvironment() returns. This Environment
is also available via dependency injection.
Prior to this change if the `bwcTest` task is run then it would create
task for each version, but each task in reality would use wireCompatVersions - 1
ES version. So we were not actually testing against 5.6.x versions in the
6.x and 6.0 branches.
This commit adjusts the assertions for the sequence number BWC tests to
account for the fact that sometimes these tests are run in
mixed-clusters with 5.6 nodes (that do not understand sequence numbers),
and sometimes these tests are run in mixed-cluster with 6.0+ nodes (that
all understood sequence numbers).
Relates #27251
Windows handles trying to read a file that does not exist because a
component of the path is not a directory differently than other OS
handle this situation. This commit adjusts these assertions for Windows.
Finder creates these files if you browse a directory there. These files
are really annoying, but it's an incredible pain for users that these
files are created unbeknownst to them, and then they get in the way of
Elasticsearch starting. This commit adds leniency on macOS only to skip
these files.
Relates #27108
Today all these API calls have a sideeffect of making documents visible
to search requests. While this is sometimes desired it's an unnecessary sideeffect
and now that we have an internal (engine-private) index reader (#26972) we artificially
add a refresh call for bwc. This change removes this sideeffect in 7.0.
The rolling-upgrade test was only writing the "minimum_master_nodes" setting to the configuration file of the old nodes, but not the upgraded ones.
Also changes the value of "minimum_master_nodes" from "number_of_nodes" to "(number_of_nodes / 2) + 1".
Today we return a `String[]` that requires copying values for every
access. Yet, we already store the setting as a list so we can also directly
return the unmodifiable list directly. This makes list / array access in settings
a much cheaper operation especially if lists are large.
The shard preference _primary, _replica and its variants were useful
for the asynchronous replication. However, with the current impl, they
are no longer useful and should be removed.
Closes#26335
The single shard optimization that we have in our search api changes the type of response returned by the query transport action name based on the shard search request. if the request goes to one shard, we will do query and fetch at the same time, hence the response will be different. The proxying layer used in cross cluster search was not aware of this distinction, which causes serialization issues every time a cross cluster search request goes to a single shard and goes through a gateway node which has to forward the shard request to a data node. The coordinating node would then expect a QueryFetchSearchResult while the gateway would return a QuerySearchResult.
Closes#26833
Since `#getAsMap` exposes internal representation we are trying to remove it
step by step. This commit is cleaning up some xcontent writing as well as
usage in tests
The ensure green approach to avoid allocation delays caused problems with other indices created by other tests which didn't use ensure green in the various cluster stages. This aligns testHistoryUUIDIsGenerated to use the same approach used by the other test.
This commit increases the amount of time to wait for green to accound for unassigned shards that
have been delayed. The default delay is 60s, so we need to wait longer than that. Previously, the
wait would timeout at 30s due to the rest client and the default for the cluster health api.
Closes#26742
The test starts with two old nodes and creates indices (without waiting for green, which is fixed here too). Then it restarts one of the nodes and waits for it to join the cluster. This wait condition only uses wait for yellow as our generic infra doesn't how many nodes are there in total. Once the restarted node is part of the cluster (mixed mode) the second old node is restarted. If indices are not fully allocated when that happens, the shards will go into delayed unassigned mode. If the recovery of the replica never completed we may end up with corrupted / no secondary copy on the node. This will cause the shards to be delayed for 1m before being reassigned and the test will time out.
It is the exciting return of the global checkpoint background
sync. Long, long ago, in snapshot version far, far away we had and only
had a global checkpoint background sync. This sync would fire
periodically and send the global checkpoint from the primary shard to
the replicas so that they could update their local knowledge of the
global checkpoint. Later in time, as we sped ahead towards finalizing
the initial version of sequence IDs, we realized that we need the global
checkpoint updates to be inline. This means that on a replication
operation, the primary shard would piggy back the global checkpoint with
the replication operation to the replicas. The replicas would update
their local knowledge of the global checkpoint and reply with their
local checkpoint. However, this could allow the global checkpoint on the
primary to advance again and the replicas would fall behind in their
local knowledge of the global checkpoint. If another replication
operation never fired, then the replicas would be permanently behind. To
account for this, we added one more sync that would fire when the
primary shard fell idle. However, this has problems:
- the shard idle timer defaults to five minutes, a long time to wait
for the replicas to learn of the new global checkpoint
- if a replica missed the sync, there was no follow-up sync to catch
them up
- there is an inherent race condition where the primary shard could
fall idle mid-operation (after having sent the replication request to
the replicas); in this case, there would never be a background sync
after the operation completes
- tying the global checkpoint sync to the idle timer was never natural
To fix this, we add two additional changes for the global checkpoint to
be synced to the replicas. The first is that we add a post-operation
sync that only fires if there are no operations in flight and there is a
lagging replica. This gives us a chance to sync the global checkpoint to
the replicas immediately after an operation so that they are always kept
up to date. The second is that we add back a global checkpoint
background sync that fires on a timer. This timer fires every thirty
seconds, and is not configurable (for simplicity). This background sync
is smarter than what we had previously in the sense that it only sends a
sync if the global checkpoint on at least one replica is lagging that of
the primary. When the timer fires, we can compare the global checkpoint
on the primary to its knowledge of the global checkpoint on the replicas
and only send a sync if there is a shard behind.
Relates #26591
The `fielddata` field and the use of the `_name` field in the short syntax of the range
query have been deprecated in 5.0 and can be removed.
The same goes for the deprecated `score_mode` field in HasParentQueryBuilder,
the deprecated `like_text`, `ids` and `docs` parameter in the `more_like_this` query,
the deprecated query name in the short version of the `regexp` query, and several
deprecated alternative field names in other query builders.
Restoring a shard from snapshot throws the primary back in time violating assumptions and bringing the validity of global checkpoints in question. To avoid problems, we should make sure that a shard that was restored will never be the source of an ops based recovery to a shard that existed before the restore. To this end we have introduced the notion of `histroy_uuid` in #26577 and required that both source and target will have the same history to allow ops based recoveries. This PR make sure that a shard gets a new uuid after restore.
As suggested by @ywelsch , I derived the creation of a `history_uuid` from the `RecoverySource` of the shard. Store recovery will only generate a uuid if it doesn't already exist (we can make this stricter when we don't need to deal with 5.x indices). Peer recovery follows the same logic (note that this is different than the approach in #26557, I went this way as it means that shards always have a history uuid after being recovered on a 6.x node and will also mean that a rolling restart is enough for old indices to step over to the new seq no model). Local shards and snapshot force the generation of a new translog uuid.
Relates #10708Closes#26544
The new ops based recovery, introduce as part of #10708, is based on the assumption that all operations below the global checkpoint known to the replica do not need to be synced with the primary. This is based on the guarantee that all ops below it are available on primary and they are equal. Under normal operations this guarantee holds. Sadly, it can be violated when a primary is restored from an old snapshot. At the point the restore primary can miss operations below the replica's global checkpoint, or even worse may have total different operations at the same spot. This PR introduces the notion of a history uuid to be able to capture the difference with the restored primary (in a follow up PR).
The History UUID is generated by a primary when it is first created and is synced to the replicas which are recovered via a file based recovery. The PR adds a requirement to ops based recovery to make sure that the history uuid of the source and the target are equal. Under normal operations, all shard copies will stay with that history uuid for the rest of the index lifetime and thus this is a noop. However, it gives us a place to guarantee we fall back to file base syncing in special events like a restore from snapshot (to be done as a follow up) and when someone calls the truncate translog command which can go wrong when combined with primary recovery (this is done in this PR).
We considered in the past to use the translog uuid for this function (i.e., sync it across copies) and thus avoid adding an extra identifier. This idea was rejected as it removes the ability to verify that a specific translog really belongs to a specific lucene index. We also feel that having a history uuid will serve us well in the future.
Boaz Leskes
Lee Hinman
Jason Tedor
David Turner
Simon Willnauer
javanna
Luca Cavanna
Michael Basnight
Nhat Nguyen
Tanguy Leroux
Clinton Gormley
Martijn van Groningen
Yannick Welsch
David Roberts
Anton Pozhidaev
kel
Nhat
Jay Modi
Christoph Büscher