Un-assign persistent tasks as nodes exit the cluster #37656
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
davidkyle
added
:Distributed Coordination/Task Management
Collaborator
|
Pinging @elastic/es-distributed |
This was referenced Jan 21, 2019
ywelsch
approved these changes
Jan 21, 2019
Contributor
There was a problem hiding this comment.
just use buildNewFakeTransportAddress() (comes from ESTestCase)
davidkyle
force-pushed
the
unassign-ptask-on-node-loss
branch
from
f48d79d to
1616eaa
Compare
davidkyle
added a commit
that referenced
this pull request
Jan 22, 2019
PersistentTasksClusterService decides if a task should be reassigned by checking there is a node in the cluster with the same Id. If a node is restarted PersistentTasksClusterService may not observe the change and decide the task still has a valid assignment because the node's ephemeral Id is not used in that decision. This change un-assigns tasks as the nodes in the cluster change.
jasontedor
added a commit
to jasontedor/elasticsearch
that referenced
this pull request
Jan 22, 2019
* elastic/master: (43 commits) Remove remaining occurances of "include_type_name=true" in docs (elastic#37646) SQL: Return Intervals in SQL format for CLI (elastic#37602) Publish to masters first (elastic#37673) Un-assign persistent tasks as nodes exit the cluster (elastic#37656) Fail start of non-data node if node has data (elastic#37347) Use cancel instead of timeout for aborting publications (elastic#37670) Follow stats api should return a 404 when requesting stats for a non existing index (elastic#37220) Remove deprecated FieldNamesFieldMapper.Builder#index (elastic#37305) Document that date math is locale independent Bootstrap a Zen2 cluster once quorum is discovered (elastic#37463) Upgrade to lucene-8.0.0-snapshot-83f9835. (elastic#37668) Mute failing test Fix java time formatters that round up (elastic#37604) Removes awaits fix as the fix is in. (elastic#37676) Mute failing test Ensure that max seq # is equal to the global checkpoint when creating ReadOnlyEngines (elastic#37426) Mute failing discovery disruption tests Add note about how the body is referenced (elastic#33935) Define constants for REST requests endpoints in tests (elastic#37610) Make prepare engine step of recovery source non-blocking (elastic#37573) ...
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
PersistentTasksClusterService decides if a task should be reassigned by checking there is a node in the cluster with the same Id instead of comparing the ephemeral Id or the actual DiscoveryNode.
https://github.com/elastic/elasticsearch/blob/master/server/src/main/java/org/elasticsearch/persistent/PersistentTasksClusterService.java#L389
During a rolling restart
PersistentTasksClusterServicemay not observe a node exit and re-enter the cluster. When the node returns the assignment appears valid as a node with the correct Id exists in the cluster but its departure and hence the un-assignment of the task was missed. This happens most often during the rolling upgrade tests where there are 3 nodes and minimum master nodes is set to 3. The solution is to un-assign the persistent tasks when the executor node leaves the cluster.There is a spelling mistake in the use of
deassociate...the correct options are disassociate or dissociate. I've followed the existing pattern for this change and will correct the spelling in a follow up PR.