Enrich store should only update the policies via an update task. #41944
Conversation
martijnvg
added
>non-issue
:Data Management/Ingest Node
|
Pinging @elastic/es-core-features |
| throw new ResourceNotFoundException("policy [{}] not found", name); | ||
| } | ||
| updateClusterState(clusterService, handler, current -> { | ||
| final Map<String, EnrichPolicy> policies = getPolicies(current); |
There was a problem hiding this comment.
should the retrieval and throw happen outside of the task? The throw will just bubble up as an exception in the caller handler passed in to EnrichStore whereas before it was a thrown exception.
There was a problem hiding this comment.
We have to check whether the policy exists in the update task, only here we can be certain about checks. We can also check it outside the update task, before we update the cluster state, but then we would check it twice.
The error will be caught by ClusterService and passed down to the handler (via ClusterStateUpdateTask's onFailure() method), so I think we're good here?
…e_under_update_thread
|
@elasticmachine run elasticsearch-ci/2 |
…e_under_update_thread
|
@elasticmachine run elasticsearch-ci/packaging-sample |
If delete or put policy requests would happen concurrently then updates may be missed without this change.