metric: add debug information to panicking goroutines in Snapshot #140266
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
Issue cockroachdb#136186 introduced a seemingly impossible situation in which the internal properties of the prometheus histogram implementation mismatched in a way which caused a panic. The properties specifically, where `h` is a histogram were `h.upperBounds` and `h.counts[0/1].buckets`. These properties are slices, initialized once, whose references are never modified, lengths never changed, and are never appended to. The length of the `counts` property is set by the length of the `upperBounds` property, which should guarantee that they are always the same length, however given the panic in the above ticket, `upperBounds` somehow became longer than `counts[0/1]`. While this change cannot tell us how the lengths of each array got to a certain way, it will tell us exactly what the lengths are, which would be the next step if it were to recur. The test in this PR is built in with a time bomb so that if it goes a year without being relevant, the test will start failing, encouraging us to move on and focus on other things. Epic: none Fixes: cockroachdb#136186 Release note: None
angles-n-daemons
requested review from
arjunmahishi and
aa-joshi
and removed request for
a team
|
This change is |
angles-n-daemons
commented
Feb 3, 2025
| cutoff := time.Date(2026, 2, 1, 0, 0, 0, 0, time.UTC) | ||
| if time.Now().After(cutoff) { | ||
| // Remove this test and the `collectHistogramBoundsCountsLengths` function. | ||
| t.Fatalf("This test has expired! It's now %v, which is after %v", time.Now(), cutoff) |
There was a problem hiding this comment.
thanks for the zest gpt..
There was a problem hiding this comment.
This is pretty cool! I haven't seen this test pattern before :)
When do we expected debugAugmentPanic to work though (release cycle wise). Will this be back-ported to older versions?
arjunmahishi
reviewed
Feb 3, 2025
| // differ. | ||
| func (h *Histogram) debugAugmentPanic() { | ||
| if r := recover(); r != nil { | ||
| stack := debug.Stack() |
There was a problem hiding this comment.
Might want to consider using this instead:
cockroach/pkg/util/debugutil/debugutil.go
Line 62 in 1eceee1
Not sure how this panic will get propagated to us (logs? sentry?). If it's through logs, it might get redacted.
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.
Issue #136186 introduced a seemingly impossible situation in which the internal properties of the prometheus histogram implementation mismatched in a way which caused a panic.
The properties specifically, where
his a histogram wereh.upperBoundsandh.counts[0/1].buckets. These properties are slices, initialized once, whose references are never modified, lengths never changed, and are never appended to. The length of thecountsproperty is set by the length of theupperBoundsproperty, which should guarantee that they are always the same length, however given the panic in the above ticket,upperBoundssomehow became longer thancounts[0/1].While this change cannot tell us how the lengths of each array got to a certain way, it will tell us exactly what the lengths are, which would be the next step if it were to recur. The test in this PR is built in with a time bomb so that if it goes a year without being relevant, the test will start failing, encouraging us to move on and focus on other things.
Epic: none
Fixes: #136186
Release note: None