Reject task on worker still starting up #21921
Merged
Conversation
This file contains 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
findepi
force-pushed
the
findepi/reject-too-early
branch
2 times, most recently
from
7c05184 to
5e4a87c
Compare
`StartupStatus.startupComplete` was set in production code, but not in tests.
When worker node is restarted after a crash, coordinator may be still unaware of the situation and may attempt to schedule tasks on it. Ideally the coordinator should not schedule tasks on worker that is not ready, but in pipelined execution there is currently no way to move a task. Accepting a request too early will likely lead to some failure and HTTP 500 (INTERNAL_SERVER_ERROR) response. The coordinator won't retry on this. Send 503 (SERVICE_UNAVAILABLE) so that request is retried.
findepi
force-pushed
the
findepi/reject-too-early
branch
from
5e4a87c to
9be0554
Compare
|
draft because of the TMP commit |
|
stress test OK https://github.com/trinodb/trino/actions/runs/9034341743/job/24827391786?pr=21921 (the actual number are probably a lie -- #20113) |
findepi
force-pushed
the
findepi/reject-too-early
branch
from
9be0554 to
8fbbd54
Compare
|
cc @sajjoseph |
|
I like the lighter version. Let me try this out with 447 release. |
|
I tried out the new PR and so far all are positive signs. Worker started processing fresh queries coming in. There were few exceptions in the worker node's server log file. I don't think these have anything to do with the current PR. I just included them here to see if any zombie tasks status checks are still causing these errors in the log files. These entries are showing up even after worker node was up for 10+ minutes. |
They look wrong cc @wendigo . i think we can assume this is not related to this PR, as it only changes anything during the startup phase. We should have an issue about them. |
|
These errors are |
Can it mean a bad thing? or should we just ignore these? |
losipiuk
approved these changes
May 13, 2024
|
@findepi I think that we should ignore. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
When worker node is restarted after a crash, coordinator may be still
unaware of the situation and may attempt to schedule tasks on it.
Ideally the coordinator should not schedule tasks on worker that is not
ready, but in pipelined execution there is currently no way to move a
task. Accepting a request too early will likely lead to some failure
and HTTP 500 (INTERNAL_SERVER_ERROR) response. The coordinator won't
retry on this. Send 503 (SERVICE_UNAVAILABLE) so that request is
retried.
Simpler alternative to #21744.
That other PR has the problem that after worker restarts queries fail for some time because the new instance ID is not propagated.
The test being added here should prove that we can close #21735.