ruler: add more details to failures for TestRulerEvaluationDelay #7217
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
Related to XXX This test is flaky and it's always flaky with the same failure - the first sample's value is 4 instead of 2. My hypothesis is that when `now.Add(-evaluationDelay)` is done, then the time range `[now.Add(-evaluationDelay), now]` might become so very slightly smaller than `evaluationDelay` when taken with the `1s` step. This results in sending a query which doesn't quite include the `5m` of samples, but `4.99999(9)m` of samples, which excludes the first sample with value `2` I couldn't get the test to fail locally, so I'm changing the assertion type to not abort the test on the first mismatch so that we can investigate the all returned samples. Also some debug logs to verify this hypothesis with the step alignment. Signed-off-by: Dimitar Dimitrov <dimitar.dimitrov@grafana.com>
gotjosh
approved these changes
Jan 25, 2024
gotjosh
reviewed
Jan 25, 2024
| @@ -293,6 +293,7 @@ func TestRulerEvaluationDelay(t *testing.T) { | |||
| require.NoError(t, mimir.WaitSumMetrics(e2e.Greater(ruleEvaluationsAfterPush[0]+float64(samplesToSend)), "cortex_prometheus_rule_evaluations_total")) | |||
|
|
|||
| // query all results to verify rules have been evaluated correctly | |||
| t.Log("querying from ", now.Add(-evaluationDelay), "to", now) | |||
There was a problem hiding this comment.
Another approach to this would be to add another assertion on the metric cortex_prometheus_last_evaluation_samples instead of cortex_prometheus_rule_evaluations_total.
This metric is incremented before the samples are pushed.
This is not the case for last_evaluation_samples.
There was a problem hiding this comment.
I suspect that the problem is that we ignore the sample even though it was pushed.
Pushed:
2, 4, 6, 8, 10, 12, 14, 16, 18, 20
Queried:
4, 6, 8, 10, 12, 14, 16, 18, 20
The failure happens because the first sample we query is 4 and not 2. TO me that fact that 4 exists means that 2 must exist as well, we just don't get it in the query results.
dimitarvdimitrov
deleted the
dimitar/ruler/TestRulerEvaluationDelay-more-failures
branch
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.
Related to #4857
This test is flaky and it's always flaky with the same failure - the first sample's value is 4 instead of 2 here:
mimir/integration/ruler_test.go
Line 314 in 9313dfc
My hypothesis is that when
now.Add(-evaluationDelay)is done the time range[now.Add(-evaluationDelay), now]may becomes so very slightly smaller thanevaluationDelay(5m) when taken with the1sstep. This results in sending a query which doesn't quite include the5mof samples, but4.99999(9)mof samples, which excludes the first sample with value2and returns4instead.I couldn't get the test to fail locally, so I'm changing the assertion type to not abort the test on the first mismatch so that we can investigate the all returned samples. Also some debug logs to verify this hypothesis with the step alignment.