Random the compression factor starting with 1 (to elimitinate nearly 0 values)
which will only use one centroid (and yield 0 for MAD as the aproximate median
is the same as the single centroid mean value)
(cherry picked from commit 940e0f1fde0f40f99af117dd03ab0891c9eedae6)
Signed-off-by: Andrei Dan <andrei.dan@elastic.co>
Summary of the issue and the root cause:
```
(1) SELECT 100, 100 -> success
(2) SELECT ?, ? (with params: 100, 100) -> success
(3) SELECT 100, 100 FROM test -> Unknown output attribute exception for the second 100
(4) SELECT ?, ? FROM test (params: 100, 100) -> Unknown output attribute exception for the second ?
(5) SELECT field1 as "x", field1 as "x" FROM test -> Unknown output attribute exception for the second "x"
```
There are two separate issues at play here:
1. Construction of `AttributeMap`s keeps only one of the `Attribute`s with the same name even if the `id`s are different (see the `AttributeMapTests` in this PR). This should be fixed no matter what, we should not overwrite attributes with one another during the construction of the `AttributeMap`.
2. The `id` on the `Alias`es is not the same in case the `Alias`es have the same `name` and same `child`
It was considered to simpy fix the second issue by just reassigning the same `id`s to the `Alias`es with the same name and child, but it would not solve the `unknown output attribute exception` (see notes below). This PR covers the fix for the first issue.
Relates to #56013
The SchedulerEngine used by SLM uses a custom runnable that will
schedule itself for its next execution if there is one to run. For the
majority of jobs, this scheduling could be many hours or days away. Due
to the scheduling so far in advance, there is a chance that time drifts
on the machine or even that time varies core to core so there is no
guarantee that the job actually runs on or after the scheduled time.
This can cause some jobs to reschedule themselves for the same
scheduled time even if they ran only a millisecond prior to the
scheduled time, which causes unexpected actions to be taken such as
what appears as duplicated snapshots.
This change resolves this by checking the triggered time against the
scheduled time and using the appropriate value to ensure that we do
not have unexpected job runs.
Relates #63754
Backport of #64501
testMoveToStepRereadsPolicy relied on an updated ILM policy that had
a rollover condition that enabled the index to be rolled after one second.
This changes the test to use a `max_doc`:1 condition so it's under the
test's control to trigger the condition.
(cherry picked from commit 73ab35a411bcdf5a92eb3d2b3bae5b1132a2bb56)
Signed-off-by: Andrei Dan <andrei.dan@elastic.co>
The yml "Test Invalid Move To Step With Invalid Next Step" worked based on
assuming the current step is a particular one. As we can't control the
timing of ILM and we can't busy assert in yml test, this converts the
test to a java test and makes use of `assertBusy`
This converts the explain lifecycle yml tests that depende on ILM having run
at least once to a java integration test that makes use of `assertBusy`.
(cherry picked from commit 6afd0422ed5ff0e3a2e5661f0e6d192bdad9af4f)
Signed-off-by: Andrei Dan <andrei.dan@elastic.co>
This field is added from version 7.11 onwards. We are
adding it to the list of ignored fields for data frame analytics
in 7.10 to avoid failing to start an outlier detection job
in a mixed cluster environment.
Relates #64503
ForecastIT.testOverflowToDisk has been observed to fail a few
times in FIPS JVMs because it takes longer than the permitted
30 seconds. This PR bumps the timeout up to 60 seconds.
Fixes#63793
This test failed sometimes for various reasons: an empty bulk request
that can't be validated, a background force-merge that completes after
the store stats were collected and finally an assertBusy() that waits
10 seconds while we usually wait 60s on the follower cluster in CCR
tests.
Closes#64167
Register a new task `runEqlCorrectnessNode` which enables developers to
start an ES node in debug mode, properly restore the correctness data
and then run queries against it.
Assert the index is restored correctly and use new snapshot.
(cherry picked from commit fc8c6dd56d602b4a62ee1ff484f00caab92dc6e2)
These tests were added to do a proper end-to-end test of the
memory usage of the geotile_grid and geohash_grid aggregations
on `geo_shape` fields. Although this was asperational,
the truth is — the test environment does not run these aggregations
in isolation. This means that the memory overhead is variable and
too flaky to rely on over time. The unit tests for circuit-breaking
remain.
Closes#63158.
This commit internalizes whether or not a role represents the ability to
contain data. In the future, this will let us remove the compatibility
role notion.
Wrap a verification_exception in case there is no valid index available in an index_not_found_exception providing also the original index pattern that may be lost in the chain of filters involving the Security one.
(cherry picked from commit 9c9da2f2f9a4ad12704f7d3a273f067e96cd2054)
The new fields option allows to fetch the value of all fields in the mapping.
However, internal fields that are used by some field mappers are also shown when
concrete fields retrieved through a pattern (`*` or `foo*`).
We have a [long term plan](https://github.com/elastic/elasticsearch/issues/63446) to hide these fields in field_caps and from pattern resolution
so this change is just a hot fix to ensure that they don't break the retrieval in the meantime.
The `flattened._keyed field will show up as an empty field when using a pattern that match the
flattened field.
Relates #63446
This commit adjusts the defaults for the tiered data roles so that they
are enabled by default, or if the node has the legacy data role. This
ensures that the default experience is that the tiered data roles are
enabled.
To fully specifiy the behavior for the tiered data roles then:
- starting a new node with the defaults: enabled
- starting a new node with node.roles configured: enabled if and only
if the tiered data roles are explicitly configured, independently
of the node having the data role
- starting a new node with node.data enabled: enabled unless the
tiered data roles are explicitly disabled
- starting a new node with node.data disabled: disabled unless the
tiered data roles are explicitly enabled
XPack usage starts out on management threads, but depending on the
implementation of the usage plugin, they could end up running on
transport threads instead. Fixed to always reschedule on a management
thread.
With this change, we will always return the same point in time in a
search response as its input until we implement the retry mechanism
for the point in times.
The `NodeNotConnectedException` exception can be nested as well in the
fairly unlikley case of the disconnect occuring between the connected check
and actually sending the request in the transport service.
Closes#63233
There is a small chance that the file deletion will run
on the searchable snapshot thread pool and not on the test
thread now that the cache is non-blocking in which case
we fail the assertion unless we wait for that thread.
Adds support for the unsigned_long type to data frame analytics.
This type is handled in the same way as the long type. Values
sent to the ML native processes are converted to floats and
hence will lose accuracy when outside the range where a float
can uniquely represent long values.
Backport of #64066
Replacing the mechanism for eviction and listener references via a read-write lock by
a reference counting implementation.
This fixes a bug that caused test failure #63586 in which concurrently trying to acquire or release
an eviction listener while doing a file operation would sometimes lead to throwing an exception
since the `tryLock` call on the read lock would fail in this case.
Also this removes the possibility of blocking cluster state updates as a result of them waiting
on the write-lock which might take a long time if a slow read operation executes concurrently.
Closes#63586
The deprecation indexing code was writing to a regular data stream,
and it is not yet possible to hide a data stream or prefix it with
a period. This functionality we be re-added once it is possible to
mark a data stream as hidden, and also to not rely on the standard
logs template since that can be disabled.
The officially supported way to clearing all entries from a cache is to use
wildcard of either * or _all. Though empty string has the same effect, it was
never intended. Therefore the tests should not use empty string and this PR
changes them to use *.
add support for unsigned_long, which required a change in
writing out integer results properly, because coerce is not
supported for unsigned_long
fixes#63871
backport #63940
When calculating feature importance, the leaf values directly correlate the value of the importance.
Consequently, positive leaf values -> positive feature importance
negative leaf values -> negative feature importance.
It follows that for binary classification, this is done such that the importance relates to the leaf values, which relate directly to the "probability of class 1".
So, the feature importance calculated is always for the importance as it relates to class 1.
The inverse is the importance as it relates to class 0.
Max and min aggs were producing wrong results for unsigned_long field
if field was indexed. If field is indexed for max/min aggs instead of
field data, we use values from indexed Points, values of which
are derived using method pointReaderIfPossible. Before
UnsignedLongFieldType#pointReaderIfPossible was incorrectly
producing values, as it failed to shift them back to original
values.
This patch fixes method pointReaderIfPossible to produce
correct original values.
Relates to #60050
* [DOCS] Combining important config settings into a single page (#63849)
* Combining important config settings into a single page.
* Updating ids for two pages causing link errors and implementing redirects.
* Updating links to use IDs instead of xrefs.
Adam Locke
Jim Ferenczi
Dan Hermann
Andrei Dan
Andras Palinkas
Jay Modi
Mayya Sharipova
Dimitris Athanasiou
Ignacio Vera
David Roberts
Tanguy Leroux
Marios Trivyzas
Tal Levy
Jason Tedor
James Rodewig
Armin Braun
Andrei Stefan
Costin Leau
Henning Andersen
Nhat Nguyen
David Kyle
Rory Hunter
Yang Wang
Hendrik Muhs
Benjamin Trent
Julie Tibshirani