Debug bootstrap for intermittent test in periodic job scheduler #271
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
brandur
force-pushed
the
brandur-debug-build-for-intermittency
branch
2 times, most recently
from
6f1d2ad to
2562ce2
Compare
This one's a special CI and logging set up aimed to help me debug #215, which I'm finding practically impossible to reproduce locally, but which happens with reasonable frequency in CI. Not intended for merge.
brandur
force-pushed
the
brandur-debug-build-for-intermittency
branch
from
2562ce2 to
0180bb2
Compare
brandur
added a commit
that referenced
this pull request
Mar 17, 2024
…ady margin" This one's aimed at fixing the intermittent test described in #215. By reading some additional logging probes in #271, I've been able to see that the failures occur when the two jobs in the test case, 500ms and 1500ms, have their ready times diverge ever so slightly across run iterations, which given enough test runs, can diverge enough so that both jobs aren't ready on the last check where they're supposed to sync up. There's a few different potential ways to solve this one, most involving rewriting the test, but here I'm proposing that we solve it by increasing the "now margin" in the enqueuer from 10ms to 100ms. This is a small margin applied on each run loop that look for jobs that are not quite ready, but so close to ready that we enqueue them anyway. The choice of 10ms originally was somewhat arbitrary, and 100ms isn't a substantially larger number and still makes sense, so I think this is a reasonable resolution. I verified the fix works by pushing it up in the repro bootstrap in #271 that runs with `-race` and high iteration count in the slower GitHub CI environment. It moves the CI matrix from failing on every job on every run to succeeding on every job on every run.
brandur
added a commit
that referenced
this pull request
Mar 17, 2024
…ady margin" This one's aimed at fixing the intermittent test described in #215. By reading some additional logging probes in #271, I've been able to see that the failures occur when the two jobs in the test case, 500ms and 1500ms, have their ready times diverge ever so slightly across run iterations, which given enough test runs, can diverge enough so that both jobs aren't ready on the last check where they're supposed to sync up. There's a few different potential ways to solve this one, most involving rewriting the test, but here I'm proposing that we solve it by increasing the "now margin" in the enqueuer from 10ms to 100ms. This is a small margin applied on each run loop that look for jobs that are not quite ready, but so close to ready that we enqueue them anyway. The choice of 10ms originally was somewhat arbitrary, and 100ms isn't a substantially larger number and still makes sense, so I think this is a reasonable resolution. I verified the fix works by pushing it up in the repro bootstrap in #271 that runs with `-race` and high iteration count in the slower GitHub CI environment. It moves the CI matrix from failing on every job on every run to succeeding on every job on every run. Fixes #215.
|
Opened #274 with a fix. |
brandur
added a commit
that referenced
this pull request
Mar 17, 2024
…ady margin" (#274) This one's aimed at fixing the intermittent test described in #215. By reading some additional logging probes in #271, I've been able to see that the failures occur when the two jobs in the test case, 500ms and 1500ms, have their ready times diverge ever so slightly across run iterations, which given enough test runs, can diverge enough so that both jobs aren't ready on the last check where they're supposed to sync up. There's a few different potential ways to solve this one, most involving rewriting the test, but here I'm proposing that we solve it by increasing the "now margin" in the enqueuer from 10ms to 100ms. This is a small margin applied on each run loop that look for jobs that are not quite ready, but so close to ready that we enqueue them anyway. The choice of 10ms originally was somewhat arbitrary, and 100ms isn't a substantially larger number and still makes sense, so I think this is a reasonable resolution. I verified the fix works by pushing it up in the repro bootstrap in #271 that runs with `-race` and high iteration count in the slower GitHub CI environment. It moves the CI matrix from failing on every job on every run to succeeding on every job on every run. Fixes #215.
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.
This one's a special CI and logging set up aimed to help me debug #215,
which I'm finding practically impossible to reproduce locally, but which
happens with reasonable frequency in CI. Not intended for merge.