sql/stats: more fully undo addOuterBuckets in stripOuterBuckets #135310
+205
−1
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 |
yuzefovich
approved these changes
Nov 15, 2024
There was a problem hiding this comment.
Nice find!
Reviewed 2 of 2 files at r1, all commit messages.
Reviewable status:complete! 1 of 0 LGTMs obtained (waiting on @michae2 and @rytaft)
pkg/sql/stats/merge.go line 109 at r1 (raw file):
// [{1, 0, 0, 2}, {1, 0, 0, 3}, {1, 0, 0, 4}] // // Partial Statistic: {row, 8, dist: 4, null: 0, size: 1}
nit: s/row, 8/row: 8/ 😃
Near the end of `stats.(*histogram).adjustCounts`, if there is leftover DistinctCount after adjusting each histogram bucket, we call `addOuterBuckets`. This function creates two "outer" buckets that fully cover the lower and upper ends of the domain with the remaining DistinctCount. These "outer" buckets interfere with partial stats, so we remove them with `stripOuterBuckets` when merging partial stats. `stripOuterBuckets` was simply slicing off the two buckets, but this could potentially leave non-zero DistinctCount in the first bucket (which is added by `addOuterBuckets`). `stripOuterBuckets` needs to also zero the DistinctCount and NumRange of the first bucket to fully undo the actions of `addOuterBuckets`. We were only catching this when trying to forecast using the merged histogram, which would sometimes produce a forecasted histogram with non-zero DistinctRange in the first bucket, leading to the infamous "histogram had first bucket with non-zero NumRange or DistinctRange" Sentry failure. Unfortunately, I don't think this bug fully explains all of the Sentry failures that look like cockroachdb#93892 because many of those predate the introduction of partial stats. I believe this specific failure started with v24.3.0-alpha.0. Informs: cockroachdb#93892 Fixes: cockroachdb#134031 (No release note because automatic collection of partial stats is not yet enabled in any available release.) Release note: None
michae2
commented
Nov 16, 2024
There was a problem hiding this comment.
TFTR!
bors r=yuzefovich
Reviewable status:
pkg/sql/stats/merge.go line 109 at r1 (raw file):
Previously, yuzefovich (Yahor Yuzefovich) wrote…
nit:
s/row, 8/row: 8/😃
lol, how did you see this? nice. done.
craig bot
pushed a commit
that referenced
this pull request
Nov 16, 2024
135310: sql/stats: more fully undo addOuterBuckets in stripOuterBuckets r=yuzefovich a=michae2 Near the end of `stats.(*histogram).adjustCounts`, if there is leftover DistinctCount after adjusting each histogram bucket, we call `addOuterBuckets`. This function creates two "outer" buckets that fully cover the lower and upper ends of the domain with the remaining DistinctCount. These "outer" buckets interfere with partial stats, so we remove them with `stripOuterBuckets` when merging partial stats. `stripOuterBuckets` was simply slicing off the two buckets, but this could potentially leave non-zero DistinctCount in the first bucket (which is added by `addOuterBuckets`). `stripOuterBuckets` needs to also zero the DistinctCount and NumRange of the first bucket to fully undo the actions of `addOuterBuckets`. We were only catching this when trying to forecast using the merged histogram, which would sometimes produce a forecasted histogram with non-zero DistinctRange in the first bucket, leading to the infamous "histogram had first bucket with non-zero NumRange or DistinctRange" Sentry failure. Unfortunately, I don't think this bug fully explains all of the Sentry failures that look like #93892 because many of those predate the introduction of partial stats. I believe this specific failure started with v24.3.0-alpha.0. Informs: #93892 Fixes: #134031 (No release note because automatic collection of partial stats is not yet enabled in any available release.) Release note: None Co-authored-by: Michael Erickson <[email protected]>
yuzefovich
reviewed
Nov 16, 2024
There was a problem hiding this comment.
Reviewed 1 of 1 files at r2, all commit messages.
Reviewable status:
pkg/sql/stats/merge.go line 109 at r1 (raw file):
Previously, michae2 (Michael Erickson) wrote…
lol, how did you see this? nice. done.
Was reading through the comments to understand what's happening in the code :)
|
Build failed: |
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.
Near the end of
stats.(*histogram).adjustCounts, if there is leftover DistinctCount after adjusting each histogram bucket, we calladdOuterBuckets. This function creates two "outer" buckets that fully cover the lower and upper ends of the domain with the remaining DistinctCount.These "outer" buckets interfere with partial stats, so we remove them with
stripOuterBucketswhen merging partial stats.stripOuterBucketswas simply slicing off the two buckets, but this could potentially leave non-zero DistinctCount in the first bucket (which is added byaddOuterBuckets).stripOuterBucketsneeds to also zero the DistinctCount and NumRange of the first bucket to fully undo the actions ofaddOuterBuckets.We were only catching this when trying to forecast using the merged histogram, which would sometimes produce a forecasted histogram with non-zero DistinctRange in the first bucket, leading to the infamous "histogram had first bucket with non-zero NumRange or DistinctRange" Sentry failure.
Unfortunately, I don't think this bug fully explains all of the Sentry failures that look like #93892 because many of those predate the introduction of partial stats. I believe this specific failure started with v24.3.0-alpha.0.
Informs: #93892
Fixes: #134031
(No release note because automatic collection of partial stats is not yet enabled in any available release.)
Release note: None