When running Elasticsearch on a flaky network, we may see nodes leaving the
cluster with reason `disconnected`. It may be useful to the cluster
administrator to see the full exception that caused the disconnection, but this
is only available with `TRACE` level logging which commingles the details of
the problem with other messages that are not useful to end users.
This commit promotes logging of exceptions in `TcpTransport` from `TRACE` to
`DEBUG` to separate them from the truly `TRACE`-level messages.
The changes are to help users prepare for migration to next major
release (v8.0.0) regarding to the break change of realm order config.
Warnings are added for when:
* A realm does not have an order config
* Multiple realms have the same order config
The warning messages are added to both deprecation API and loggings.
The main reasons for doing this are: 1) there is currently no automatic relay
between the two; 2) deprecation API is under basic and we need logging
for OSS.
This commit switches the strategy for managing dot-prefixed indices that
should be hidden indices from using "fake" system indices to an explicit
exclusions list that must be updated when those indices are converted to
hidden indices.
* Optimize not-equalities in con-/disjunctions
This commit adds optimisations of not-equalities in conjunctions and
disjunctions:
* for conjunctions, the not-equality can be optimized away when applied
together with a range or inequality, in case the not-equality point
falls outside the domain of the later condition; if its on the boarder,
it will modify the bound, to simply exclude the equality, if present;
otherwise no optimisation can be applied;
* for disjunctions, the not-equals could filter away the ranges and
inequalities, unless these include an equality on the bound, in which
case the entire condition becomes always true, but this would influence
the score() function, so it's been omitted;
* fix aggregations of inequalities in ranges
This commit fixes the loop that aggregates inequalities into ranges:
- it won't advance the outer loop index in case of a merge, since the
current element is removed;
- it will break the inner loop, since comparision against the element
selected in the outer loop can't continue, as it had been removed.
(cherry picked from commit 789724ac2cc726de603849b4eeb8194da7528bcc)
* Rename ILM history index enablement setting
The previous setting was `index.lifecycle.history_index_enabled`, this commit changes it to
`indices.lifecycle.history_index_enabled` to indicate this is not an index-level setting (it's node
level).
* [ML][Inference] Fix weighted mode definition (#51648)
Weighted mode inaccurately assumed that the "max value" of the input values would be the maximum class value. This does not make sense.
Weighted Mode should know how many classes there are. Hence the new parameter `num_classes`. This indicates what the maximum class value to be expected.
testRecovery relies on the fact that shards are not flushed on inactive.
Our CI recently was too slow. It took more than 20 minutes to complete
the full cluster restart suite. This slowness caused some shards of
testRecovery were flushed on inactive.
This commit increases the inactive time to 1h to reduce this noise.
Closes#51640
We need to force flush to make the last commit safe; otherwise, we might
fail to open FrozenEngine. Note that we force flush before closing a
shard.
Closes#51620
Three fixes for when the `compressed_definition` is utilized on PUT
* Update the inflate byte limit to be the minimum of 10% the max heap, or 1GB (what it was previously)
* Stream data directly to the JSON parser, so if it is invalid, we don't have to inflate the whole stream to find out
* Throw when the maximum bytes are reach indicating that is why the request was rejected
* Fix InternalEngineTests.testSeqNoAndCheckpoints
If we force flush while possibly triggering a merge the local checkpoint may change
from the expectation from the loop that just increments on every operation.
Closes#51604
* [DOCS] Rewrite analysis intro. Move index/search analysis content.
* Rewrites 'Text analysis' page intro as high-level definition.
Adds guidance on when users should configure text analysis
* Rewrites and splits index/search analysis content:
* Conceptual content -> 'Index and search analysis' under 'Concepts'
* Task-based content -> 'Specify an analyzer' under 'Configure...'
* Adds detailed examples for when to use the same index/search analyzer
and when not.
* Adds new example snippets for specifying search analyzers
* clarifications
* Add toc. Decrement headings.
* Reword 'When to configure' section
* Remove sentence from tip
Previously, if YEAR() was used as and ORDER BY argument without being
wrapped with another scalar (e.g. YEAR(birth_date) + 10), no script
ordering was used but instead the underlying field (e.g. birth_date)
was used instead as a performance optimisation. This works correctly if
YEAR() is the only ORDER BY arg but if further args are used as tie
breakers for the ordering wrong results are produced. This is because
2 rows with the different birth_date but on the same year are not tied
as the underlying ordering is on birth_date and not on the
YEAR(birth_date), and the following ORDER BY args are ignored.
Remove this optimisation for YEAR() to avoid incorrect results in
such cases.
As a consequence another bug is revealed: scalar functions on top
of nested fields produce scripted sorting/filtering which is not yet
supported. In such cases no error was thrown but instead all values for
such nested fields were null and were passed to the script implementing
the sorting/filtering, producing incorrect results.
Detect such cases and throw a validation exception.
Fixes: #51224
(cherry picked from commit f41efd6753dc3650a7eabb3e07b02b3b32c5704c)
There is no reason not to allow deletes in parallel to restores
if they're dealing with different snapshots.
A delete will not remove any files related to the snapshot that
is being restored if it is different from the deleted snapshot
because those files will still be referenced by the restoring
snapshot.
Loading RepositoryData concurrently to modifying it is concurrency
safe nowadays as well since the repo generation is tracked in the
cluster state.
Closes#41463
set watcher logger to debug level.
These tests haven't run in such a long time,
we first need to get a better picture how/if
these tests fail today.
Backport of #51478
See #33185
Add a verification that full-text search functions `MATCH()` and `QUERY()`
are not allowed in the SELECT clause, so that a nice error message is
returned to the user early instead of an "ugly" exception.
Fixes: #47446
ActionListener.map would call listener.onFailure for exceptions from
listener.onResponse, but this means we could double trigger some
listeners which is generally unexpected. Instead, we should assume that
a listener's onResponse (and onFailure) implementation is responsible
for its own exception handling.
The audit index is re-created for every testrun and therefore potential useful debug information
gets lost. This change reads out the audit index and logs the results, which makes them available
for debugging CI issues.
relates #51549
This commit creates a new index privilege named `maintenance`.
The privilege grants the following actions: `refresh`, `flush` (also synced-`flush`),
and `force-merge`. Previously the actions were only under the `manage` privilege
which in some situations was too permissive.
Co-authored-by: Amir H Movahed <arhd83@gmail.com>
Datafeeds being closed while starting could result in and NPE. This was
handled as any other failure, masking out the NPE. However, this
conflicts with the changes in #50886.
Related to #50886 and #51302
In #50259 we redirected stdout and stderr to log4j, to capture jdk
and external library messages. However, a typo in the method name used
to redirect the stream in java means stdout is currently being
duplicated twice, and stderr not captured. This commit corrects that
mistake. Unfortunately this is at a level that cannot really be tested,
thus we are still missing tests for this behavior.
This PR tries to address the intermittent vector test failures on 7.x by making
sure we create indices with one shard.
The fix is based on this theory as to what's happening:
* On 7.x, the default number of shards is 1, but in REST tests we randomly use
2 in order to cover the multiple shards case. In the failing test run, we use 2
shards and all documents end up on only one shard.
* During a search, the response from the empty shard doesn't produce
deprecation warnings because we never try to execute the script. If not all
shard responses contain the warning headers, then certain deprecation warnings
can be lost (due to the bug described in #33936).
Addresses #50716.
Relates to #50061.
For small uploads (that can still be up to 5MB!) we needlessly
reading the `InputStream` into a BAOS which entailed allocating
the `byte[]` for the stream contents twice (because to `toByteArray` on the BAOS copies).
Also, for resumeable uploads we were needlessly wrapping the output channel and running each individual write in its own privileged context when we could just wrap the whole upload in a single privileged context.
Relates #51593
Dominic Page
Adrien Grand
Sven Schliesing
Ioannis Kakavas
Henning Andersen
David Turner
Yang Wang
Mark Vieira
Lisa Cawley
Gordon Brown
Bogdan Pintea
Lee Hinman
Benjamin Trent
Nhat Nguyen
James Rodewig
Armin Braun
Marios Trivyzas
Martijn van Groningen
Hendrik Muhs
Albert Zaharovits
Peter Dyson
Ryan Ernst
Julie Tibshirani
Aleksandr Maus
Przemysław Witek
Przemko Robakowski
David Roberts