The .watcher-history-* template is currently using a plugin-custom index setting xpack.watcher.template.version,
which prevents this template from being installed in a mixed OSS / X-Pack cluster, ultimately
leading to the situation where an X-Pack node is constantly spamming an OSS master with (failed)
template updates. Other X-Pack templates (e.g. security-index-template or security_audit_log)
achieve the same versioning functionality by using a custom _meta field in the mapping instead.
This commit switches the .watcher-history-* template to use the _meta field instead.
Enables a rolling restart from the OSS distribution to the x-pack based distribution by preventing
x-pack code from installing custom metadata into the cluster state until all nodes are capable of
deserializing this metadata.
This commit reduces the Windows CLI scripts to one-liners by moving all
of the redundant logic to an elasticsearch-cli script. This commit is
only the Windows side, a previous commit covered the Linux side.
This commit reduces the Linux CLI scripts to one-liners by moving all of
the redundant logic to an elasticsearch-cli script. This commit is only
the Linux side, a follow-up will do this for Windows too.
Watcher tests now always fail hard when watches that were
tried to be triggered in a test using the trigger() method,
but could not because they were not found on any of the
nodes in the cluster.
This commit removes xpack from being a meta-plugin-as-a-module.
It also fixes a couple tests which were missing task dependencies, which
failed once the gradle execution order changed.
* Refactors ClientHelper to combine header logic
This change removes all the `*ClientHelper` classes which were
repeating logic between plugins and instead adds
`ClientHelper.executeWithHeaders()` and
`ClientHelper.executeWithHeadersAsync()` methods to centralise the
logic for executing requests with stored security headers.
* Removes Watcher headers constant
When the encrpytion of sensitive date is enabled, test that a
scheduled watch is executed as expected and produces the correct value
from a secret in the basic auth header.
Since adding back the per-watch statistics, we do not need to access
every trigger engine implementation to get the current total job count.
This commit removes the unused methods to do so.
The HTTPClient used in watcher is based on the apache http client. The
current client is using a lot of defaults - which are not always
optimal. Two of those defaults are the maximum number of total
connections and the maximum number of connections to a single route.
If one of those limits is reached, the HTTPClient waits for a connection
to be finished thus acting in a blocking fashion. In order to prevent
this when many requests are being executed, we increase the limit of
total connections as well as the connections per route (a route is
basically an endpoint, which also contains proxy information, not
containing an URL, just hosts).
On top of that an additional option has been set to evict
long running connections, which can potentially be reused after some
time. As this requires an additional background thread, this required
some changes to ensure that the httpclient is closed properly. Also the
timeout for this can be configured.
Starting watcher should wait for the watcher to be started before
marking the status as started, which is now done via a callback.
Also, reloading watcher could set the execution service to paused. This could
lead to watches not being executed, when run in tests. This fix does not
change the paused flag in the execution service, just clears out the
current queue and executions.
Closes#30381
When the watcher service pauses execution due to a cluster state update,
the trigger service and its engines also need to pause properly instead
of keeping going. This is also important when the .watches index is
deleted, so that watches don't stay in a triggered mode.
The current implementation starts/stops watcher using an executor. This
can result in our of order operations.
This commit reduces those executor calls to an absolute minimum in order
to be able to do state changes within the cluster state listener method,
which runs in sequence.
When a state change occurs that forces the watcher service to pause
(like no watcher index, no master node, no local shards), the service is
now in a paused state.
Pausing is a super lightweight operation, which marks the
ExecutionService as paused and waits for the currently executing watches
to finish in the background via an executor. The same applies for
stopping, the potentially long running operation is outsourced in to an
executor, as waiting for executed watches is decoupled from the current
state.
The only other long running operation is starting, where watches need to
be loaded. This is also done via an executor, but has an additional
protection by checking the cluster state version it was started with. If
another cluster state version was trying to load the watches, then this
loading will not take effect.
This PR also cleans up some unused states, like the a simple boolean in
the HistoryStore/TriggeredWatchStore marking it as started or stopped,
as this can now be caught in the execution service.
Another advantage of this approach is the fact, that now only triggered
watches are not getting executed, while watches that are run via the
Execute Watch API will still be executed regardless if watcher is
stopped or not.
Lastly the TickerScheduleTriggerEngine thread now only starts on data nodes.
This commit removes the http.enabled setting. While all real nodes (started with bin/elasticsearch) will always have an http binding, there are many tests that rely on the quickness of not actually needing to bind to 2 ports. For this case, the MockHttpTransport.TestPlugin provides a dummy http transport implementation which is used by default in ESIntegTestCase.
closes#12792
The variadic constructor was only used in a few places and the
RepositoriesMetaData class is backed by a List anyway, so just using a
List will make it simpler to instantiate it.
Starting with the refactoring in https://github.com/elastic/elasticsearch/pull/22778 (released in 5.3) we may fail to properly replicate operation when a mapping update on master fails. If a bulk
operations needs a mapping update half way, it will send a request to the master before continuing
to index the operations. If that request times out or isn't acked (i.e., even one node in the cluster
didn't process it within 30s), we end up throwing the exception and aborting the entire bulk. This is
a problem because all operations that were processed so far are not replicated any more to the
replicas. Although these operations were never "acked" to the user (we threw an error) it cause the
local checkpoint on the replicas to lag (on 6.x) and the primary and replica to diverge.
This PR does a couple of things:
1) Most importantly, treat *any* mapping update failure as a document level failure, meaning only
the relevant indexing operation will fail.
2) Removes the mapping update callbacks from `IndexShard.applyIndexOperationOnPrimary` and
similar methods for simpler execution. We don't use exceptions any more when a mapping
update was successful.
I think we need to do more work here (the fact that a single slow node can prevent those mappings
updates from being acked and thus fail operations is bad), but I want to keep this as small as I can
(it is already too big).
We had a number of awaitsFix links that weren't updated after the xpack
merge.
Where possible I changed the links to the new locations, but in some
circumstances the original ticket was closed (suggesting the awaitsfix
should be removed) or was otherwise unclear the status.
Email message IDs are supposed to be unique. In order to guarantee this,
we need to take the action id of a watch action into account as well,
not just the watch id from the watch execution context. This prevents
that two actions from the same watch execution end up with the same
message id.
This commit adds the distribution type to the startup scripts so that we
can discern from log output and the main response the type of the
distribution (deb/rpm/tar/zip).
This commit adds the distribution flavor (default versus oss) to the
build process which is passed through the startup scripts to
Elasticsearch. This change will be used to customize the message on
attempting to install/remove x-pack based on the distribution flavor.
This commit makes x-pack a module and adds it to the default
distrubtion. It also creates distributions for zip, tar, deb and rpm
which contain only oss code.
Igor Motov
Yannick Welsch
Jason Tedor
Alexander Reelsen
Ryan Ernst
Colin Goodheart-Smithe
Michael Basnight
Boaz Leskes
Zachary Tong