Commit Graph

28738 Commits

Author SHA1 Message Date
Simon Willnauer b7bc790428 Use a non default port range in MockTransportService
We already use a per JVM port range in MockTransportService. Yet,
it's possible that if we are executing in the JVM with ordinal 0 that
other clusters reuse ports from the mock transport service and some tests
try to simulate disconnects etc. By using a non-defautl port range (starting at 10300)
we prevent internal test clusters from reusing any of the mock impls ports

Relates to #25301
2017-07-12 22:29:21 +02:00
Simon Willnauer e81804cfa4 Add a shard filter search phase to pre-filter shards based on query rewriting (#25658)
Today if we search across a large amount of shards we hit every shard. Yet, it's quite
common to search across an index pattern for time based indices but filtering will exclude
all results outside a certain time range ie. `now-3d`. While the search can potentially hit
hundreds of shards the majority of the shards might yield 0 results since there is not document
that is within this date range. Kibana for instance does this regularly but used `_field_stats`
to optimize the indexes they need to query. Now with the deprecation of `_field_stats` and it's upcoming removal a single dashboard in kibana can potentially turn into searches hitting hundreds or thousands of shards and that can easily cause search rejections even though the most of the requests are very likely super cheap and only need a query rewriting to early terminate with 0 results.

This change adds a pre-filter phase for searches that can, if the number of shards are higher than a the `pre_filter_shard_size` threshold (defaults to 128 shards), fan out to the shards
and check if the query can potentially match any documents at all. While false positives are possible, a negative response means that no matches are possible. These requests are not subject to rejection and can greatly reduce the number of shards a request needs to hit. The approach here is preferable to the kibana approach with field stats since it correctly handles aliases and uses the correct threadpools to execute these requests. Further it's completely transparent to the user and improves scalability of elasticsearch in general on large clusters.
2017-07-12 22:19:20 +02:00
Jason Tedor 86e9438d3c Prevent excessive disk consumption by log files
This commit enables management of the main Elasticsearch log files
out-of-the-box by the following changes:
 - compress rolled logs
 - roll logs every 128 MB
 - maintain a sliding window of logs
 - remove the oldest logs maintaining no more than 2 GB of compressed
   logs on disk

Relates #25660
2017-07-12 15:52:00 -04:00
Luca Cavanna 8b846f9141 Migrate RestHttpResponseHeadersIT to ESRestTestCase (#25675)
Closes #25611
2017-07-12 21:42:50 +02:00
Jason Tedor 5a416b9922 Use config directory to find jvm.options
This commit removes the environment variable ES_JVM_OPTIONS that allows
the jvm.options file to sit separately from the rest of the config
directory. Instead, we use the CONF_DIR environment variable for custom
configuration location just as we do for the other configuration files.

Relates #25679
2017-07-12 15:29:13 -04:00
Jason Tedor 39b94b72b3 Fix inadvertent rename of systemd tests
This commit reverts a rename of the systemd packaging tests. The rename
was done locally to speed up iteration of testing some changes against
systemd but was not reverted before pushing. This commit reverts this
change.
2017-07-12 15:02:51 -04:00
Christoph Büscher f3e7a1c4a4 Adding basic search request documentation for high level client (#25651) 2017-07-12 17:06:46 +02:00
Jack Conradson d2b4f7ac5a Disallow lang to be used with Stored Scripts (#25610)
Requests that execute a stored script will no longer be allowed to specify the lang of the script. This information is stored in the cluster state making only an id necessary to execute against. Putting a stored script will still require a lang.
2017-07-12 07:55:57 -07:00
Antonio Matarrese 8d7cbc43b5 Fix typo in ScriptDocValues deprecation warnings (#25672) 2017-07-12 16:17:55 +02:00
Colin Goodheart-Smithe 55a157e964 Changes DocValueFieldsFetchSubPhase to reuse doc values iterators for multiple hits (#25644)
* Changes DocValueFieldsFetchSubPhase to reuse doc values iterators for multiple hits

Closes #24986

* iter

* Update ScriptDocValues to not reuse GeoPoint and Date objects

* added Javadoc about script value re-use
2017-07-12 12:03:49 +00:00
Martijn van Groningen 0a25558f98
Query range fields by doc values when they are expected to be more efficient than points.
* Enable doc values for range fields by default.
* Store ranges in a binary format that support multi field fields.
* Added BinaryDocValuesRangeQuery that can query ranges that have been encoded into a binary doc values field.
* Wrap range queries on a range field in IndexOrDocValuesQuery query.

Closes #24314
2017-07-12 13:04:14 +02:00
Christoph Büscher ad01a67c51 Remove SearchHit#internalHits (#25653)
This method does exactly what getHits() does and is used in only a few places,
so it can safely be removed. It seems to be a left-over from when
InternalSearchHits was folded into the SearchHits interface, which didn't
contain this method.
2017-07-12 10:01:18 +02:00
Deb Adair b5e81132cf [DOCS] Reorganized the highlighting topic so it's less confusing. 2017-07-11 21:16:14 -07:00
Jason Tedor e165c405ac Add an underscore to flood stage setting
This is a minor nitty bikeshedding change that renames the suffix of the
disk flood stage setting to "flood_stage" from "floodstage".

Relates #25659
2017-07-11 22:02:00 -04:00
Jason Tedor 3a827827c1 Avoid failing install if system-sysctl is masked
On Debian-based systems the install scripts are run with set -e meaning
that if there is an error in executing one of these scripts then the
script fails. If systemd-sysctl is masked then trying to restart the
systemd-sysctl service to pick up the changes to vm.max_map_count will
fail leading to the post-install script failing. Instead, we should
account for the possbility of failure here by not letting the command to
restart this service exit with non-zero status code. This commit does
this, and adds a test for this situation.

Relates #25657
2017-07-11 17:38:50 -04:00
James Baiera 847378a43b Add another parent value option to join documentation (#25609)
Indexing a join field on a document requires a value of type "object" and two sub fields "name" 
and "parent". The "parent" field is only required on child documents, but the "name" field which 
denotes the name of the relation is always needed. Previously, only the short-hand version of the 
join field was documented. This adds documentation for the long-hand join field data, and 
explicitly points out that just specifying the name of the relation for the field value is a 
convenience shortcut.
2017-07-11 15:36:59 -04:00
Simon Willnauer 831dbbf291 Ensure we rewrite common queries to `match_none` if possible (#25650)
In certain situations we can early terminate and just skip the entire
query phase or make the lucene level rewrite very cheap if we can already
tell that a query won't match any documents. For instance if there is a single
`match_none` ie. due to some range rewrite in a filter or must clause of a boolean
query it can just drop all it's other queries since it will never match.
2017-07-11 21:19:14 +02:00
Adrien Grand de99610c4e Remove reference to field-stats docs. 2017-07-11 18:38:25 +02:00
Adrien Grand f9fbce84b6 Optimize the order of bytes in uuids for better compression. (#24615)
Flake ids organize bytes in such a way that ids are ordered. However, we do not
need that property and could reorganize bytes in an order that would better suit
Lucene's terms dict instead.

Some synthetic tests suggest that this change decreases the disk footprint of
the `_id` field by about 50% in many cases (see `UUIDTests.testCompression`).
For instance, when simulating the indexing of 10M docs at a rate of 10k docs
per second, the current uid generator used 20.2 bytes per document on average,
while this new generator which only puts bytes in a different order uses 9.6
bytes per document on average.

We had already explored this idea in #18209 but the attempt to share long common
prefixes had had a bad impact on indexing speed. This time I have been more
careful about putting discriminant bytes early in the `_id` in a way that
preserves indexing speed on par with today, while still allowing for better
compression.
2017-07-11 17:28:23 +02:00
Tim Brooks a3ade99fcf Fix BytesReferenceStreamInput#skip with offset (#25634)
There is a bug when a call to `BytesReferenceStreamInput` skip is made
on a `BytesReference` that has an initial offset. The offset for the
current slice is added to the current index and then subtracted from the
length. This introduces the possibility of a negative number of bytes to
skip. This happens inside a loop, which leads to an infinte loop.

This commit correctly subtracts the current slice index from the
slice.length. Additionally, the `BytesArrayTests` are modified to test
instances that include an offset.
2017-07-11 09:54:29 -05:00
Simon Willnauer 98c91a3bd0 Limit the number of concurrent shard requests per search request (#25632)
This is a protection mechanism to prevent a single search request from
hitting a large number of shards in the cluster concurrently. If a search is
executed against all indices in the cluster this can easily overload the cluster
causing rejections etc. which is not necessarily desirable. Instead this PR adds
a per request limit of `max_concurrent_shard_requests` that throttles the number of
concurrent initial phase requests to `256` by default. This limit can be increased per request
and protects single search requests from overloading the cluster. Subsequent PRs can introduces
addiontional improvemetns ie. limiting this on a `_msearch` level, making defaults a factor of
the number of nodes or sort shards iters such that we gain the best concurrency across nodes.
2017-07-11 16:23:10 +02:00
Clinton Gormley bd7ddfa175 Removed field-stats docs 2017-07-11 15:15:25 +02:00
Adrien Grand 481d5d09b2 Upgrade to lucene-7.0.0-snapshot-00142c9. (#25641)
Lucene 7.0 is feature-frozen now, so there should not be many changes until GA.
2017-07-11 13:58:55 +02:00
Clinton Gormley 92849c64db Fixed bad asciidoc file name 2017-07-11 12:47:52 +02:00
Clinton Gormley ddbbe9f7cc Tidied up the breaking changes docs 2017-07-11 12:40:14 +02:00
Simon Willnauer 538110bd60 Change compatibility version to 5.6 after backport 2017-07-11 11:39:08 +02:00
Simon Willnauer ec1afe30ea Ensure remote cluster alias is preserved in inner hits aggs (#25627)
We lost the cluster alias due to some special caseing in inner hits
and due to the fact that we didn't pass on the alias to the shard request.
This change ensures that we have the cluster alias present on the shard to
ensure all SearchShardTarget reads preserve the alias.

Relates to #25606
2017-07-11 11:34:06 +02:00
Herman Schaaf 977712f977 Change small typo in shards_allocation.asciidoc (#25643) 2017-07-11 11:25:49 +02:00
Tal Levy e04be73ad5 remove ingest.new_date_format (#25583) 2017-07-10 13:07:50 -07:00
Tim Brooks b22bbf94da Avoid blocking on channel close on network thread (#25521)
Currently when we close a channel in Netty4Utils.closeChannels we
block until the closing is complete. This introduces the possibility
that a network selector thread will block while waiting until a
separate network selector thread closes a channel.

For instance: T1 closes channel 1 (which is assigned to a T1 selector).
Channel 1's close listener executes the closing of the node. That
means that T1 now tries to close channel 2. However, channel 2 is
assigned to a selector that is running on T2. T1 now must wait until T2
closes that channel at some point in the future.

This commit addresses this by adding a boolean to closeChannels
indicating if we should block on close. We only set this boolean to true
if we are closing down the server channels at shutdown. This call is
never made from a network thread. When we call the closeChannels method
with that boolean set to false, we do not block on close.
2017-07-10 10:50:51 -05:00
Tal Levy 8cf0528001 update ingest-user-agent regexes.yml (#25608)
This new regexes are from:

3153c2f2ae/regexes.yaml
2017-07-10 08:43:11 -07:00
Yannick Welsch 7836bbf4d4 Fix tribe node cluster state version increments (#25629)
With #24236, tribe nodes submit cluster state changes to their MasterService, making it unnecessary to explicitly update the cluster state version. This PR fixes the double-incrementing of cluster state versions on tribe nodes, which are not harmful, but unnecessary.
2017-07-10 16:25:11 +02:00
Colin Goodheart-Smithe 3a5a54e83e Collapses package structure for some bucket aggs (#25579)
This change collapses some of the packages for the bucket aggregations into their parent packages. This was done for the following aggregations:
* The variants of the range aggregation (geo_distance, date and ip) were moved into the `o.e.s.a.bucket.range` package
* The `o.e.s.a.bucket.terms.support` package was removed and the classes were moved to `o.e.s.a.bucket.terms`
* The filter aggregation was moved to `o.e.s.a.bucket.filter`

Since this PR is already relatively large with only the above changes subsequent PRs will do similar operations on relevant metric and pipeline aggregations

Relates to #22868
2017-07-10 15:08:15 +01:00
Boaz Leskes e93e10f93b Close Translog trimming task when IndexService is closed
Relates to https://github.com/elastic/elasticsearch/pull/25622
2017-07-10 14:40:23 +02:00
Yannick Welsch b5521872bb [TEST] Use correct StreamInput version to deserialize in testSnapshotDeletionsInProgressSerialization
The test is currently serializing the cluster state using an older ES version format, but then deserializes those same bytes by
assuming they are of the current ES version.
2017-07-10 14:03:12 +02:00
Clinton Gormley e85871cfe9 Update cross-cluster-search.asciidoc
Increased the required min version of CCS in the docs to 5.5
2017-07-10 12:04:05 +02:00
Boaz Leskes a6db0ea908 Run Translog retention yaml tests with no replicas
Initializing replicas change the translog retention logic and confuses the test.

Switch to the solution suggested in https://github.com/elastic/elasticsearch/issues/25623, if implemented
2017-07-10 11:06:57 +02:00
Luca Cavanna a932591007 Treat aliases as unavailable indices in delete index and update aliases api (#25524)
When resolving wildcards, aliases should be treated as unavailable indices when the `ignoreAliases` option is set to `true` (currently enabled with delete index api and update aliases api). This way the `allow_no_indices` and `ignore_unavailable` options can be honoured, otherwise WildcardExpressionResolver ends up treating aliases differently and there is no way to control when an error is thrown.

The default behaviour for the delete index api, which has `ignore_unavailable` set to `false` and `allow_no_indices` set to `true` by default, is to throw an error when executed against an alias, same as when it's executed against an index that does not exist.
2017-07-10 10:58:00 +02:00
Boaz Leskes 09378f48e4 Add a scheduled translog retention check (#25622)
We currently check whether translog files can be trimmed whenever we create a new translog generation or close a view. However #25294 added a long translog retention period (12h, max 512MB by default), which means translog files should potentially be cleaned up long after there isn't any indexing activity to trigger flushes/the creation of new translog files. We therefore need a scheduled background check to clean up those files once they are no longer needed.

Relates to #10708
2017-07-10 10:28:39 +02:00
Jason Tedor c084542731 Bump version to 6.0.0-beta1
This commit does two things:
 - bumps the version from 6.0.0-alpha3 to 6.0.0-beta1
 - renames the 6.0.0-alpha3 version constant to 6.0.0-beta1

Relates #25621
2017-07-09 18:12:50 -04:00
Jason Tedor c75ddd2c85 Fix scaling thread pool test bug
This commit adjusts the expectation for the max number of threads in the
scaling thread pool configuration test. The reason that this expectation
is incorrect is because we removed the limitation that the number of
processors maxes out at 32, instead letting it be the true number of
logical processors on the machine. However, when we removed this
limitation, this test was never adjusted to reflect the new reality yet
it never arose since our tests were not running on machines with
incredibly high core counts.

Relates #20874
2017-07-09 08:00:27 -04:00
Boaz Leskes 1f4d8a05d1 testConcurrentWriteViewsAndSnapshot: writers should expose the local checkpoint to readers before trimming the translog 2017-07-09 12:26:54 +02:00
Jason Tedor cb3674c5ee Add reason to global checkpoint updates on replica
Updating the global checkpoint on a replica can occur for a few
different reasons:
 - from inlined global checkpoint updates
 - from a primary term transition
 - from finalizing recovery

Yet, the trace logging for a global checkpoint update does not present
this information that can be useful when tracing test failures. This
commit adds a reason for the global checkpoint update on a replica so
that we can trace these updates.

Relates #25612
2017-07-08 17:05:24 -04:00
Jason Tedor fe04b666ea Add console appender to deprecated settings test
This commit enables the console appender for the settings deprecation
logger used in the deprecated settings logging test. This output will be
useful if this test fails again (it failed once mysteriously).
2017-07-08 07:59:35 -04:00
Boaz Leskes 215bffb08b Enable bwc testing
#25512 & #25511 have been merged
2017-07-08 11:57:22 +02:00
Boaz Leskes 40ae134f5a Move `BulkItemRequest` BWC to 5.x (#25511)
The current BWC code in `BulkItemRequest` mutates the underlying `DocWriteRequests` which causes test failures and unexpected state (our test infra checks bwc serialization on the fly). This PR removes this logic from master. Another PR will add a BWC layer to 5.x only.

This PR contains the logic in https://github.com/elastic/elasticsearch/pull/25510 , which is needed to run the tests.
2017-07-08 11:42:57 +02:00
Boaz Leskes f189e819be testRecoveryAfterPrimaryPromotion: seqNo recovery doesn't require some initial indexing
Previously the primary didn't update it's own local checkpoint (and thus the global checkpoint) before some indexing occurred. With recent changes the primary now properly initializes it self and thus ops recovery is possible even if no indexing has occurred.
2017-07-08 10:05:05 +02:00
Boaz Leskes b40b002c1d Disable bwc tests in preparation of merging #25512 & #25511 2017-07-08 09:25:43 +02:00
DeDe Morton baa1858f56 [DOCS] Fix link (#25616) 2017-07-07 20:40:44 -07:00
DeDe Morton a4fedb213e Fix attribute reference on redirects page (#25614) 2017-07-07 20:15:42 -07:00