When a watch is acknowledged, while it is also being executed, the
acknowledgment information can get lost. The reason for this is the
fact, that the execution writes the watch status inside of the watch
regardless, if other writes happened inbetween to make sure the
execution state is caught.
This commit checks the current executions in the execution service and
aborts the API call, if the specified watch ID can be found in those.
Note, this does not prevent this issue fully, as a watch could be
triggered, while the acknowledgement update is running, but it does
reduce the surface area of this problem. In order to properly solve
this, indexing the watch status as part of a watch would need to be
changed.
relates elastic/x-pack-elasticsearch#4003
Original commit: elastic/x-pack-elasticsearch@d7e218b2ac
... yet support updates. This commit introduces a few changes of how
watches are put.
The GET Watch API will never return credentials like basic auth
passwords, but a placeholder instead now. If the watcher is enabled to
encrypt sensitive settings, then the original encrypted value is
returned otherwise a "::es_redacted::" place holder.
There have been several Put Watch API changes.
The API now internally uses the Update API and versioning. This has
several implications. First if no version is supplied, we assume an
initial creation. This will work as before, however if a credential is
marked as redacted we will reject storing the watch, so users do not
accidentally store the wrong watch.
The watch xcontent parser now has an additional methods to tell the
caller if redacted passwords have been found. Based on this information
an error can be thrown.
If the user now wants to store a watch that contains a password marked
as redacted, this password will not be part of the toXContent
representation of the watch and in combinatination with update request
the existing password will be merged in. If the encrypted password is
supplied this one will be stored.
The serialization for GetWatchResponse/PutWatchRequest has changed.
The version checks for this will be put into the 6.x branch.
The Watcher UI now needs specify the version, when it wants to store a
watch. This also prevents last-write-wins scenarios and is the reason
why the put/get watch response now contains the internal version.
relates elastic/x-pack-elasticsearch#3089
Original commit: elastic/x-pack-elasticsearch@bb63be9f79
The execution state is kind of a global indicator if a watch has been
running successfully and is used by the watcher UI.
However this field is only stored in the watch history but not part of
the watch status, thus it is not available everywhere. In order to
simplify the watcher UI this commit also adds the field to the
watch status which is stored together with the watch.
It is stored under the `status.execution_state` field as `status.state`
is already taken. This is also reflects with the name of the java class.
The WatchStatus class does not contain serialization checks, as this is
intended to be backported to 6.x, where those checks will be added.
Once the backport is done, the old execution state field can be fully
deleted from the master branch in another commit (syncing with Kibana
folks required).
relates elastic/x-pack-elasticsearch#2385
* fix doc tests
Original commit: elastic/x-pack-elasticsearch@26e8f99571
* [DOCS] Format Watcher APIs
* [DOCS] Removed master_timeout from Watcher APIs
* [DOCS] Added authority info to watcher APIs
Original commit: elastic/x-pack-elasticsearch@1e6de3b036
As there are no master node operations anymore.
* TransportActions are regular Actions now
* Watcher requests are now ActionRequests, no MasterNodeRequests anymore
* REST spec does not contain master node timeout parameters anymore
* WatcherLifeCycleService does not have a check anymore if watcher is able to run distributed, this will be a given in 7.0
* Some serialization BWC checks against version 5 have been removed
Original commit: elastic/x-pack-elasticsearch@4607dd538c
The deprecated handlers should have been removed earlier, but are now
going to to away finally.
Also the watcher restart action has been removed, mainly because users
should not blindly restart, but always make sure, that watcher is
stopped correctly before restarting. This had been removed from the
transport action previously.
Original commit: elastic/x-pack-elasticsearch@78a5ec3c05
As fields with underscores will be disallowed in master, and we have to
prepare the upgrade, this commit renames the _status field to status.
When the 5.x upgrade logic is in place in the 5.x we can remove all the
old style _status handling from the master branch.
Note: All the BWC compatibility tests, that load 5.x indices are now
faking a finished upgrade by adding the `status` field to the mapping
of the watches index.
Original commit: elastic/x-pack-elasticsearch@9d5cc9aaec
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
Alexander Reelsen
Lisa Cawley