Backport of (#47721) for 7.x.
Similarly to #47582, Auto-follow patterns creates following
indices as long as the remote index matches the pattern and
the remote primary shards are all started. But since 7.2 closed
indices are also replicated, and it does not play well with CCR
auto-follow patterns as they create following indices for closed
leader indices too.
This commit changes the getLeaderIndicesToFollow() so that closed
indices are excluded from auto-follow patterns.
If a cluster sending monitoring data is unhealthy and triggers an
alert, then stops sending data the following exception [1] can occur.
This exception stops the current Watch and the behavior is actually
correct in part due to the exception. Simply fixing the exception
introduces some incorrect behavior. Now that the Watch does not
error in the this case, it will result in an incorrectly "resolved"
alert. The fix here is two parts a) fix the exception b) fix the
following incorrect behavior.
a) fixing the exception is as easy as checking the size of the
array before accessing it.
b) fixing the following incorrect behavior is a bit more intrusive
- Note - the UI depends on the success/met state for each condition
to determine an "OK" or "FIRING"
In this scenario, where an unhealthy cluster triggers an alert and
then goes silent, it should keep "FIRING" until it hears back that
the cluster is green. To keep the Watch "FIRING" either the index
action or the email action needs to fire. Since the Watch is neither
a "new" alert or a "resolved" alert, we do not want to keep sending
an email (that would be non-passive too). Without completely changing
the logic of how an alert is resolved allowing the index action to
take place would result in the alert being resolved. Since we can
not keep "FIRING" either the email or index action (since we don't
want to resolve the alert nor re-write the logic for alert resolution),
we will introduce a 3rd action. A logging action that WILL fire when
the cluster is unhealthy. Specifically will fire when there is an
unresolved alert and it can not find the cluster state.
This logging action is logged at debug, so it should be noticed much.
This logging action serves as an 'anchor' for the UI to keep the state
in an a "FIRING" status until the alert is resolved.
This presents a possible scenario where a cluster starts firing,
then goes completely silent forever, the Watch will be "FIRING"
forever. This is an edge case that already exists in some scenarios
and requires manual intervention to remove that Watch.
This changes changes to use a template-like method to populate the
version_created for the default monitoring watches. The version is
set to 7.5 since that is where this is first introduced.
Fixes#43184
This commit adds documentation for new index privilege
create_doc which only allows indexing of new documents
but no updates to existing documents via Index or Bulk APIs.
Relates: #45806
* Separate SLM stop/start/status API from ILM
This separates a start/stop/status API for SLM from being tied to ILM's
operation mode. These APIs look like:
```
POST /_slm/stop
POST /_slm/start
GET /_slm/status
```
This allows administrators to have fine-grained control over preventing
periodic snapshots and deletions while performing cluster maintenance.
Relates to #43663
* Allow going from RUNNING to STOPPED
* Align with the OperationMode rules
* Fix slmStopping method
* Make OperationModeUpdateTask constructor private
* Wipe snapshots better in test
Failed snapshots will eventually build up unless they are deleted. While
failures may not take up much space, they add noise to the list of
snapshots and it's desirable to remove them when they are no longer
useful.
With this change, failed snapshots are deleted using the following
strategy: `FAILED` snapshots will be kept until the configured
`expire_after` period has passed, if present, and then be deleted. If
there is no configured `expire_after` in the retention policy, then they
will be deleted if there is at least one more recent successful snapshot
from this policy (as they may otherwise be useful for troubleshooting
purposes). Failed snapshots are not counted towards either `min_count`
or `max_count`.
When exceptions could be returned from another node, the exception
might be wrapped in a `RemoteTransportException`. In places where
we handled specific exceptions using `instanceof` we ought to unwrap
the cause first.
This commit attempts to fix this issue after searching code in the ML
plugin.
Backport of #47676
* Convert RunTask to use testclusers, remove ClusterFormationTasks
This PR adds a new RunTask and a way for it to start a
testclusters cluster out of band and block on it to replace
the old RunTask that used ClusterFormationTasks.
With this we can now remove ClusterFormationTasks.
Previously when retrieving an SLM policy it would always return a 200
with `{}` in the body, even if the policy did not exist. This changes
that behavior to throw an error (similar to our other APIs) if a
policy doesn't exist.
This also adds a basic CRUD yml test for the behavior.
Resolves#47664
If a thread pool rejection exception happens, an alternative code
path is chosen to write history and delete the trigger. If an exception
happens during deletion of the trigger an exception may be thrown and not
caught.
This commit catches the exception and provides a meaning error message.
fixes#47008
When deactivating a watch, there is a chance that it is fully deactivated
and reporting as not running but the history is not fully written yet.
There is not a tight coupling between the associated watcher history
index and the deactivation. This test assumes that once a watch is
deactivated that all history is fully written in a very short time period.
If the Watch is deactivated, but the history is slow to write it can result
in a failing test.
This change removes an assertion that assumes that the deactivation of a watch
ensured the all of the watch history was written. There is still a minor race
condition with respect to the remaining history assertions. However, if the
history is slow to be written, it will allow the test to still passing.
fixes#47503
Adds the following parameters to `outlier_detection`:
- `compute_feature_influence` (boolean): whether to compute or not
feature influence scores
- `outlier_fraction` (double): the proportion of the data set assumed
to be outlying prior to running outlier detection
- `standardization_enabled` (boolean): whether to apply standardization
to the feature values
Backport of #47600
Previously, we supported only the format `{fn <FUNCTION_NAME>()}`
but other DBs like MSSQL, DB2, MariaDB/MySQL alos allow whitespaces
between `{` and `fn`. Furhermore, also some applications - like PowerBI -
generate escape sequences with spaces: `select { fn name(params) } etc.`
Add support for white spaces between `{` and the escape pattern definition
like `fn`, `ts`, `d`, `guid` etc.
Closes: #47401
(cherry picked from commit 08a22d0b393f4a76c52dabc5e7b9cafcc19c30ca)
Use case:
User with `create_doc` index privilege will be allowed to only index new documents
either via Index API or Bulk API.
There are two cases that we need to think:
- **User indexing a new document without specifying an Id.**
For this ES auto generates an Id and now ES version 7.5.0 onwards defaults to `op_type` `create` we just need to authorize on the `op_type`.
- **User indexing a new document with an Id.**
This is problematic as we do not know whether a document with Id exists or not.
If the `op_type` is `create` then we can assume the user is trying to add a document, if it exists it is going to throw an error from the index engine.
Given these both cases, we can safely authorize based on the `op_type` value. If the value is `create` then the user with `create_doc` privilege is authorized to index new documents.
In the `AuthorizationService` when authorizing a bulk request, we check the implied action.
This code changes that to append the `:op_type/index` or `:op_type/create`
to indicate the implied index action.
This commit adds support to retrieve all API keys if the authenticated
user is authorized to do so.
This removes the restriction of specifying one of the
parameters (like id, name, username and/or realm name)
when the `owner` is set to `false`.
Closes#46887
Backport of (#47582)
Today when following a new leader index, we fetch the remote cluster state,
check the remote cluster license, check the user privileges, retrieve the
index shard stats before initiating a CCR restore session.
But if the leader index to follow is closed, we're executing a bunch of
operations that would inevitability fail at some point (on retrieving the
index shard stats, because this type of request forbid closed indices
when resolving indices). We could fail a Put Follow request at the first
step by checking the leader index state directly from the remote cluster
state.
This also helps the Resume Follow API to fail a bit earlier.
* Use versions specific distribution folders so we don't need to clean up (#46539)
* Retry deleting distro dir on windows
When retarting the cluster we clean up old distribution files that might
still be in use by the OS.
Windows closes resources of ded processes async, so we do a couple of
retries to get arround it.
Closes#46014
* Avoid having to delete the distro folder.
* Remove the use of ClusterFormationTasks form RestTestTask (#47022)
This PR removes a use-case of the ClusterFormationTasks and converts a
project that flew under the radar so far.
There's probably more clean-up possible here, but for now the goal is
to be able to remove that code after `RunTask` is also updated.
* Migrate some 7.x only projects
An index with an ILM policy that has a rollover action in one of the
phases was rolled over when the ILM conditions dictated regardless if
it was already rolled over (eg. manually after modifying an index
template in order to force the creation of a new index that uses the new
mappings).
This changes this behaviour and has ILM check if the index it's about to
roll has not been rolled over in the meantime.
(cherry picked from commit 37d6106feeb9f9369519117c88a9e7e30f3ac797)
Signed-off-by: Andrei Dan <andrei.dan@elastic.co>
This commit lifts the validation of the monitoring hosts setting into
the setting itself, rather than when the setting is used. This prevents
a scenario where an invalid value for the setting is accepted, but then
later fails while applying a cluster state with the invalid setting.
This adds a default for the `slm.retention_schedule` setting, setting it
to `0 30 1 * * ?` which is 1:30am every day.
Having retention unset meant that it would never be invoked and clean up
snapshots. We determined it would be better to have a default than never
to be run. When coming to a decision, we weighed the option of an
absolute time (such as 1:30am) versus a periodic invocation (like every
12 hours). In the end we decided on the absolute time because it has
better predictability and consistency than a periodic invocation, which
would rely on when the master node were elected or restarted.
Relates to #43663
* Bwc testclusters all (#46265)
Convert all bwc projects to testclusters
* Fix bwc versions config
* WIP fix rolling upgrade
* Fix bwc tests on old versions
* Fix rolling upgrade
When an ML job runs the memory required can be
broken down into:
1. Memory required to load the executable code
2. Instrumented model memory
3. Other memory used by the job's main process or
ancilliary processes that is not instrumented
Previously we added a simple fixed overhead to
account for 1 and 3. This was 100MB for anomaly
detection jobs (large because of the completely
uninstrumented categorization function and
normalize process), and 20MB for data frame
analytics jobs.
However, this was an oversimplification because
the executable code only needs to be loaded once
per machine. Also the 100MB overhead for anomaly
detection jobs was probably too high in most cases
because categorization and normalization don't use
_that_ much memory.
This PR therefore changes the calculation of memory
requirements as follows:
1. A per-node overhead of 30MB for _only_ the first
job of any type to be run on a given node - this
is to account for loading the executable code
2. The established model memory (if applicable) or
model memory limit of the job
3. A per-job overhead of 10MB for anomaly detection
jobs and 5MB for data frame analytics jobs, to
account for the uninstrumented memory usage
This change will enable more jobs to be run on the
same node. It will be particularly beneficial when
there are a large number of small jobs. It will
have less of an effect when there are a small number
of large jobs.
When API key is invalidated we do two things first it tries to trigger `ExpiredApiKeysRemover` task
and second, we do index the invalidation for the API key. The index invalidation may happen
before the `ExpiredApiKeysRemover` task is run and in that case, the API key
invalidated will also get deleted. If the `ExpiredApiKeysRemover` runs before the
API key invalidation is indexed then the API key is not deleted and will be
deleted in the future run.
This behavior was not captured in the tests related to `ExpiredApiKeysRemover`
causing intermittent failures.
This commit fixes those tests by checking if the API key invalidated is reported
back when we get API keys after invalidation and perform the checks based on that.
Closes#41747
The changes introduced in #47179 made it so that we could try to
build an SSLContext with verification mode set to None, which is
not allowed in FIPS 140 JVMs. This commit address that
* Remove eclipse conditionals
We used to have some meta projects with a `-test` prefix because
historically eclipse could not distinguish between test and main
source-sets and could only use a single classpath.
This is no longer the case for the past few Eclipse versions.
This PR adds the necessary configuration to correctly categorize source
folders and libraries.
With this change eclipse can import projects, and the visibility rules
are correct e.x. auto compete doesn't offer classes from test code or
`testCompile` dependencies when editing classes in `main`.
Unfortunately the cyclic dependency detection in Eclipse doesn't seem to
take the difference between test and non test source sets into account,
but since we are checking this in Gradle anyhow, it's safe to set to
`warning` in the settings. Unfortunately there is no setting to ignore
it.
This might cause problems when building since Eclipse will probably not
know the right order to build things in so more wirk might be necesarry.
* Add API to execute SLM retention on-demand (#47405)
This is a backport of #47405
This commit adds the `/_slm/_execute_retention` API endpoint. This
endpoint kicks off SLM retention and then returns immediately.
This in particular allows us to run retention without scheduling it
(for entirely manual invocation) or perform a one-off cleanup.
This commit also includes HLRC for the new API, and fixes an issue
in SLMSnapshotBlockingIntegTests where retention invoked prior to the
test completing could resurrect an index the internal test cluster
cleanup had already deleted.
Resolves#46508
Relates to #43663
* Fix AllocationRoutedStepTests.testConditionMetOnlyOneCopyAllocated
These tests were using randomly generated includes/excludes/requires for
routing, however, it was possible to generate mutually exclusive
allocation settings (about 1 out of 50,000 times for my runs).
This splits the test into three different tests, and removes the
randomization (it doesn't add anything to the testing here) to fix the
issue.
Resolves#47142
Fixes multiple Active Directory related tests that run against the
samba fixture. Some were failing since we changed the realm settings
format in 7.0 and a few were slightly broken in other ways.
We can move to cleanup the tests in a follow up but this work fits
better to be done with or after we move the tests from a Samba
based fixture to a real(-ish) Microsoft Active Directory based
fixture.
Resolves: #33425, #35738
While it seemed like the PUT data frame analytics action did not
have to be a master node action as the config is stored in an index
rather than the cluster state, there are other subtle nuances which
make it worthwhile to convert it. In particular, it helps maintain
order of execution for put actions which are anyhow user driven and
are expected to have low volume.
This commit converts `TransportPutDataFrameAnalyticsAction` from
a handled transport action to a master node action.
Note this means that the action might fail in a mixed cluster
but as the API is still experimental and not widely used there will
be few moments more suitable to make this change than now.
Bulk requests currently do not allow adding "create" actions with auto-generated IDs.
This commit allows using the optype CREATE for append-only indexing operations. This is
mainly the user facing aspect of it.
XPackPlugin holds data in statics and can only be initialized once. This
caused tests to fail primarily when running with a low max-workers.
Replaced usages with the LocalStateCompositeXPackPlugin, which handles
this properly for testing.
Due to #47003 many clusters will have built up a
large backlog of expired results. On upgrading to
a version where that bug is fixed users could find
that the first ML daily maintenance task deletes
a very large amount of documents.
This change introduces throttling to the
delete-by-query that the ML daily maintenance uses
to delete expired results to limit it to deleting an
average 200 documents per second. (There is no
throttling for state/forecast documents as these
are expected to be lower volume.)
Additionally a rough time limit of 8 hours is applied
to the whole delete expired data action. (This is only
rough as it won't stop part way through a single
operation - it only checks the timeout between
operations.)
Relates #47103
This commit restores the model state if available in data
frame analytics jobs.
In addition, this changes the start API so that a stopped job
can be restarted. As we now store the progress in the state index
when the task is stopped, we can use it to determine what state
the job was in when it got stopped.
Note that in order to be able to distinguish between a job
that runs for the first time and another that is restarting,
we ensure reindexing progress is reported to be at least 1
for a running task.
Gordon Brown
Tanguy Leroux
Jim Ferenczi
Jake Landis
Yogesh Gaikwad
Andrei Stefan
Lee Hinman
Dimitris Athanasiou
Alpar Torok
Benjamin Trent
Hendrik Muhs
Lisa Cawley
Marios Trivyzas
Andrei Dan
Ioannis Kakavas
Jason Tedor
Przemysław Witek
David Roberts
Yannick Welsch
Henning Andersen