Fix concurrent harvesters #2541
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
ruflin
added
bug
review
Filebeat
ruflin
force-pushed
the
paralallel-harvesters-fix
branch
from
6818c40
to
b90eabd
Compare
| ok := p.outlet.OnEvent(event) | ||
| if !ok { | ||
| logp.Info("Prospector outlet closed") | ||
| return |
There was a problem hiding this comment.
change in logic here. This return used to stop this worker. The return having moved to updateState will not stop the worker potentially publishing more incoming events. The worker instead should be dropped + report to prospector it needs to 'shutdown' + drain the harvesterChan, so no harvester will be blocked.
There was a problem hiding this comment.
I added an bool to updateState to be returned so the loop can check if event was sent and return accordingly. Same is needed for other calls of updateState.
|
LGTM |
In case newly started harvesters did not persist their first state before the next scan started, it could have happened that multiple harvesters were started for the same file. This could have been cause by a large number of files or the output blocking. The problem is solve that the Setup step of the Harvester is now synchronus and blocking the scan. Part of this is also updating the first state of the as part of the prospector. The side affect of this change is that now a scan is blocking in case the channel is blocked which means the output is probably not responding. If the output is not responding, scans will not continue and new files will not be discovered until output is available again. The code can be further simplified in the future by merging create/startHarvester. This will be done in a second step to keep backport commit to a minimum. See also elastic#2539
ruflin
force-pushed
the
paralallel-harvesters-fix
branch
from
cde1400
to
dc12cfe
Compare
ruflin
force-pushed
the
paralallel-harvesters-fix
branch
from
dc12cfe
to
3086818
Compare
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.
In case newly started harvesters did not persist their first state before the next scan started, it could have happened that multiple harvesters were started for the same file. This could have been cause by a large number of files or the output blocking.
The problem is solve that the Setup step of the Harvester is now synchronus and blocking the scan. Part of this is also updating the first state of the as part of the prospector.
The side affect of this change is that now a scan is blocking in case the channel is blocked which means the output is probably not responding. If the output is not responding, scans will not continue and new files will not be discovered until output is available again.
The code can be further simplified in the future by merging create/startHarvester. This will be done in a second step to keep backport commit to a minimum.
See also #2539