-
Notifications
You must be signed in to change notification settings - Fork 163
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix(vllm metrics): error stack trace #3200
fix(vllm metrics): error stack trace #3200
Conversation
2e62654
to
9f81631
Compare
LGTM. |
9f81631
to
54a09d6
Compare
Codecov ReportAll modified and coverable lines are covered by tests ✅
Additional details and impacted files@@ Coverage Diff @@
## main #3200 +/- ##
==========================================
- Coverage 85.11% 85.08% -0.03%
==========================================
Files 1291 1291
Lines 28782 28788 +6
Branches 7744 7752 +8
==========================================
- Hits 24497 24495 -2
- Misses 4285 4293 +8
... and 5 files with indirect coverage changes Continue to review full report in Codecov by Sentry.
|
faab5b4
to
8df986a
Compare
Signed-off-by: gitdallas <[email protected]>
8df986a
to
64f493b
Compare
tested locally, works fine. |
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: mturley The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
closes: https://issues.redhat.com/browse/RHOAIENG-11522
this is a situation that would result in the error stack and ui crash without this change:
Description
prevent ui from crashing. let the query be undefined if it doesn't exist, which will result in an empty data and no errors. vince said he did not want an error message at all as it might convey to the user that it might resolve with a refresh or something.
How Has This Been Tested?
tested code on a previous deploy that would crash the ui on metrics page, it no longer crashes. used MR cluster to test. existing tests still pass.
Test Impact
added a new test using mock data that only contains 1 query and made sure that the 4 charts show up (instead of an error stack page). i also updated the test mock for
prometheus/serving
to return empty results if the request body includesquery=undefined\b
as it would in the real endpoint. Here's a screenshot from a test with a missing query resulting in no data for one of the serving endpoints (it still shows the data):Request review criteria:
test a vllm deploy, view the metrics. also view metrics of other types.
Self checklist (all need to be checked):
If you have UI changes:
After the PR is posted & before it merges:
main