block-builder: Add an option to consume a partition completely till cycle end offset with no partially region #10506
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
…ilder Signed-off-by: Ganesh Vernekar <[email protected]>
Signed-off-by: Ganesh Vernekar <[email protected]>
codesome
changed the title
Signed-off-by: Ganesh Vernekar <[email protected]>
Comment on lines
-892
to
-899
| bucketDir := path.Join(cfg.BlocksStorage.Bucket.Filesystem.Directory, "1") | ||
| db, err := tsdb.Open(bucketDir, promslog.NewNopLogger(), nil, nil, nil) | ||
| require.NoError(t, err) | ||
| t.Cleanup(func() { require.NoError(t, db.Close()) }) | ||
| compareQuery(t, | ||
| db, | ||
| producedSamples, | ||
| nil, |
There was a problem hiding this comment.
Cool, you had mentioned this, but I didn't quite see the blatant duplication. 👍
seizethedave
approved these changes
Jan 23, 2025
| // We will process this sample in the next cycle. | ||
| allSamplesProcessed = false | ||
| continue | ||
| } | ||
| if recordAlreadyProcessed && s.TimestampMs < lastBlockMax { | ||
| if !processEverything && recordAlreadyProcessed && s.TimestampMs < lastBlockMax { |
There was a problem hiding this comment.
As I understand it the first time we restart blockbuilders with this flag enabled, we will produce a small amount of duplicate records. (records that had recordAlreadyProcessed from the previous cycle, but we're no longer skipping past those.) I do not see an avenue where we'd skip any records on the first run. Am I right?
There was a problem hiding this comment.
Yes, migrating from old the new way will create some duplicate samples in blocks which is harmless since compactor will merge them together.
codesome
added a commit
that referenced
this pull request
Jan 24, 2025
…ycle end offset with no partially region (#10506) (#10514) * Add an option to not leave any partially consumed region for block builder Signed-off-by: Ganesh Vernekar <[email protected]> * Add TestNoPartiallyConsumedRegions Signed-off-by: Ganesh Vernekar <[email protected]> * Fix imports Signed-off-by: Ganesh Vernekar <[email protected]> --------- Signed-off-by: Ganesh Vernekar <[email protected]> (cherry picked from commit 5ff8293) Co-authored-by: Ganesh Vernekar <[email protected]>
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.
What this PR does
Currently we have "last seen" and "commit" points, and the records between can be partially consumed. This PR adds an experimental
block-builder.no-partially-consumed-regionwhere we do not have this partially consumed region, instead entire records until cycle end time goes inside a block.When flag is enabled, we also reduce the buffer time from 15 mins to 5 mins. Which means, the consumption will technically be from
X:05to(X+1):05. The extra 5 mins is to allow for any remote write delay. Hard coded for now for experiments.This flag is an intermediate state, meaning after running tests with this option, we will only keep one of them and get rid of the flag. The wish is to keep this new version since it makes scheduler design and parallelisation easier. If this one wins, we can also get rid of "last seen offset" from the commit information.
Checklist
[ ] Documentation added.[ ]CHANGELOG.mdupdated - the order of entries should be[CHANGE],[FEATURE],[ENHANCEMENT],[BUGFIX].about-versioning.mdupdated with experimental features.