The distribution of watches now happens on the node which holds the
watches index, instead of on the master node. This requires several
changes to the current implementation.
1. Running on shards and replicas
In order to run watches on the nodes with the watches index on its
primaries and replicas. To ensure that watches do not run twice, there is
a logic which checks the local shards, runs a murmurhash on the id and
runs modulo against the number of shards and replicas, this is the way to
find out, if a watch should run local. Reloading happens
2. Several master node actions moved to a HandledTransportAction, as they
are basically just aliases for indexing actions, among them the
put/delete/get watch actions, the acknowledgement action, the de/activate
actions
3. Stats action moved to a broadcast node action, because we potentially
have to query every node to get watcher statistics
4. Starting/Stopping watcher now is a master node action, which updates
the cluster state and then listeners acts on those. Because of this watches
can be running on two systems, if you those have different cluster state
versions, until the new watcher state is propagated
5. Watcher is started on all nodes now. With the exception of the ticker
schedule engine most classes do not need a lot of resources while running.
However they have to run, because of the execute watch API, which can hit
any node - it does not make sense to find the right shard for this watch
and only then execute (as this also has to work with a watch, that has not
been stored before)
6. By using a indexing operation listener, each storing of a watch now
parses the watch first and only stores on successful parsing
7. Execute watch API now uses the watcher threadpool for execution
8. Getting the number of watches for the stats now simply queries the
different execution engines, how many watches are scheduled, so this is
not doing a search anymore
There will be follow up commits on this one, mainly to ensure BWC compatibility.
Original commit: elastic/x-pack-elasticsearch@0adb46e658
Within the same JVM, setting the number of processors available to Netty
can only be done once. However, tests randomize the number of processors
and so without intervention would attempt to set this value multiple
times. Therefore, we need to use a flag that prevents setting this value
in tests.
Relates elastic/x-pack-elasticsearch#1266
Original commit: elastic/x-pack-elasticsearch@d127149725
Many of the tests assume that the trial license has already been generated before the test gets to run. As this is asynchronously triggered upon node
startup, however, there is no guarantee that trial license generation has completed before the tests get to execute, leading to null values when
checking clusterService.state().metaData().custom(LicensesMetaData.TYPE).
Original commit: elastic/x-pack-elasticsearch@d909c9ba95
LicenseManagerServiceTests sometimes fails in jenkins, but fairly
rarely. We don't get useful logs when it does. This cranks up
the log level and adds some more assertions so we can better track
down where the failure comes from.
Relates to elastic/x-pack-elasticsearch#222
Original commit: elastic/x-pack-elasticsearch@3e08725fc7
When a user creates a datafeed, as well as checking they have permission
to create a datafeed we also check that they have permission to search the
indices they've configured the datafeed to search.
Previously this second check was erroneously done for the user who issued
the put_datafeed request, whereas it should be done as the runas user for
that request.
Original commit: elastic/x-pack-elasticsearch@4c35204c66
When we revert to snapshot, if we delete intervening results
we should delete with an open end on the time range for the
case when future data has been posted to the job.
Original commit: elastic/x-pack-elasticsearch@c3f5e8f19e
The one argument ctor for `Script` creates a script with the
default language but most usages of are for testing and either
don't care about the language or are for use with
`MockScriptEngine`. This replaces most usages of the one argument
ctor on `Script` with calls to `ESTestCase#mockScript` to make
it clear that the tests don't need the default scripting language.
Original commit: elastic/x-pack-elasticsearch@c1d05b7357
When the active directory realm was refactored to add support for authenticating against multiple
domains, only the default authenticator respected the user_search.filter setting. This commit moves
this down to the base authenticator and also changes the UPN filter to not include sAMAccountName
in the filter.
Original commit: elastic/x-pack-elasticsearch@d2c19c9bee
- Removes need to handle exception from action methods
- Clearly renames DatafeedJobIT to DatafeedJobsRestIT to distinguish
from DatafeedJobsIT
- Refactors DatafeedJobsIT to reuse MlNativeAutodetectIntegTestCase
Original commit: elastic/x-pack-elasticsearch@5bd0c01391
Cross cluster search uses ClusterSearchShardsAction under the covers.
Without this change, you would need both "read_cross_cluster" and "view_index_metadata" privilegs in order to have permission to execute searches from a remote cluster.
Original commit: elastic/x-pack-elasticsearch@65a6aff329
If a single permission set does not have a query defined then this should be considered as the user
not having document level security for the indices matching that pattern. However, the lack of
document level security was not being taken into account and document level security was being
applied when it should not have been.
Original commit: elastic/x-pack-elasticsearch@f5777c2f37
Many tests in monitoring use the pattern of calling first awaitMonitoringDocsCount, and then doing a search that checks certain properties, assuming
that the doc count is correct at that point. In the presence of replicas, awaitMonitoringDocsCount might not wait for all shard copies to have the
desired property. A subsequent search might then hit a shard where the property does not hold. As these tests randomize the number of replicas
(through the random_index_template), it easier to constrain awaitMonitoringDocsCount just to the primary and then do subsequent checks just by
querying the primary.
Original commit: elastic/x-pack-elasticsearch@4165beb903
This commit creates a single server socket that will be connected to by local sockets. The local
sockets will use the port of the previously stopped ldap server as the local port. This will
prevent the ldap library from establishing a connection. The previous use of server sockets for
this did not work on all operating systems as the backlog parameter has platform specific meaning.
Original commit: elastic/x-pack-elasticsearch@03b6bf39d4
This commit adds a timeout for the expiration of invalidated tokens so that we can expect that the
requests will have been finished before we do the assertions on the internal test cluster.
Original commit: elastic/x-pack-elasticsearch@2928706224
This commit adds a token based access mechanism that is a subset of the OAuth 2.0 protocol. The
token mechanism takes the same values as a OAuth 2 standard (defined in RFC 6749 and RFC 6750),
but differs in that we use XContent for the body instead of form encoded values. Additionally, this
PR provides a mechanism for expiration of a token; this can be used to implement logout
functionality that prevents the token from being used again.
The actual tokens are encrypted using AES-GCM, which also provides authentication. The key for
encryption is derived from a salt value and a passphrase that is stored on each node in the
secure settings store. By default, the tokens have an expiration time of 20 minutes and is
configurable up to a maximum of one hour.
Relates elastic/x-pack-elasticsearch#8
Original commit: elastic/x-pack-elasticsearch@3d201ac2bf
This is an issue where a bucket can have both interim results and
non-interim results, a bucket should never have both at the same time.
The steps to cause this situation are:
1. Flush a running job and create interim results
2. Close that job (this does not delete interim results)
3. Re-open the job and POST data
4. The job will eventually emit a bucket result which mingles with the
existing interim results
Normally interim results are deleted by AutoDetectResultProcessor when a
bucket is parsed following a flush command. Because of the close and
re-opening of the job AutoDetectResultProcessor no longer knows that a
previous flush command creating interim results.
The fix is to always delete interim results the first time
AutoDetectResultProcessor sees a bucket.
relates elastic/x-pack-elasticsearch#1188
Original commit: elastic/x-pack-elasticsearch@5326455f54
In the SessionFactoryLoadBalancingTests, we sometime want a connection to a certain IP and Port to
fail as a way to mock an unresponsive/disconnected LDAP server. The test does this by starting up
multiple LDAP servers and then shutting some down. When the server is shut down the port that it
was bound to is open for another process or test to bind to, which can lead to sporadic failures in
CI. This change is a best effort attempt to prevent this by binding a server socket to the port and
filling its backlog so other connections should fail.
Relates elastic/x-pack-elasticsearch#1195
Original commit: elastic/x-pack-elasticsearch@b31a560c93
The DatafeedJobsIT.testRealtime test fails from time to time.
The test seems to take a long time to execute the flush action
after the lookback. This could make sense as the test produces
a few records over the span of a week with 5 minutes bucket_span.
Thus, flush will end up doing a lot of word to create results
for so many buckets.
This change increases the bucket_span to 1 hour. Hopefully, this
will stop the failures.
Relates elastic/x-pack-elasticsearch#1162
Original commit: elastic/x-pack-elasticsearch@4366907371
David Kyle
Alexander Reelsen
Jason Tedor
Yannick Welsch
Dimitrios Athanasiou
Nik Everett
David Roberts
Zachary Tong
Jay Modi
Tim Vernum
jaymode
Martijn van Groningen