backupccl: buffer/sort returned files before flushing to remote #66876
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
|
This change is |
dt
force-pushed
the
backup-sort
branch
2 times, most recently
from
3abf924 to
6a6c5ae
Compare
pbardea
reviewed
Jun 25, 2021
Release note: none.
BACKUP has multiple workers sending export requests concurrently so the returned files may arrive out of order. Previously this would always force the writer that flushed returned files out to the remote file to close the file it had been writing an open a new one, as any one file must be in-order. This adds a small queue of returned files to the sink to which it adds files it is given to write to remote storage. Once the queue has accumulated, it is sorted and partially drained to the remote file. This should increase the odds that files are added to the remote file in-order and thus do not require closing and re-opening additional remote files. With default workers on a tpcc5k cluster running an incremental backup this was observed to reduce the number of files a node wrote from ~50-70 to <10. Release note: none.
pbardea
approved these changes
Jun 28, 2021
|
TFTR! bors r+ |
|
Build succeeded: |
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Outdated suggestions cannot be applied.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.
BACKUP has multiple workers sending export requests concurrently so the
returned files may arrive out of order. Previously this would always
force the writer that flushed returned files out to the remote file to
close the file it had been writing an open a new one, as any one file
must be in-order. This adds a small queue of returned files to the sink
to which it adds files it is given to write to remote storage. Once the
queue has accumulated, it is sorted and partially drained to the remote
file. This should increase the odds that files are added to the remote
file in-order and thus do not require closing and re-opening additional
remote files.
With default worker (3) on a tpcc5k cluster running an incremental backup
this was observed to reduce the number of files a node wrote from ~50-70 to
<10.
Release note: none.