Description

plz12345

opened on Feb 10, 2026

Issue body actions

We're using the native/non-YAML CodeQL implementation. GitHub had an outage earlier today, where git itself was returning HTTP 500's for about an hour. During this time, the CodeQL jobs for our repo failed. Unlike other jobs, we have no option to retry them. It's "view" or "delete". The only workaround I could think of is closing the triggering PR and opening a new one, but I shouldn't have to do that on native GitHub functionality failing.

Reacted by Damyan Petev, Javier Padilla Keymole, John Osborn, Andreas Johansen and Steffen Schroeder

Reacted by Steffen Schroeder

Activity

1. mbg commented on Feb 10, 2026

   

   mbg

   on Feb 10, 2026

   Member

   More actions

   Hi @plz12345,

   Thanks for opening this issue and sorry about the impact of the outage on CodeQL analyses.

   I believe that the relevant team at GitHub is already aware of this limitation with managed workflows (it's not specific to CodeQL / Default Setup), but I can't say if and when that will be addressed.

   On our end (CodeQL Action) we are investigating whether we can improve when and how often the CodeQL Action retries API requests, but that would be local to steps within a workflow and probably would not have helped with an incident of this duration.

   The only workaround I could think of is closing the triggering PR and opening a new one, but I shouldn't have to do that on native GitHub functionality failing.

   Pushing e.g. an empty (or no-op) commit (or similar) should also trigger the analysis to run again, but I agree that none of these are ideal workarounds.
2. plz12345 commented on Feb 10, 2026

   

   plz12345

   on Feb 10, 2026

   Author

   More actions

   Yeah, I thought it might but this was on a protected branch,
3. nanotaboada commented on Mar 13, 2026

   

   nanotaboada

   on Mar 13, 2026

   More actions

   Two things that helped us:

   1. Workaround to re-trigger the failing job

   If you have another PR pending against the same base branch, merge it first. Then go back to the affected PR — it will now show "This branch is out-of-date with the base branch" with an Update branch button. Clicking it is enough to trigger a new push event and re-run the CodeQL checks.

   2. Migrate to CodeQL Advanced (YAML-based setup)

   The native setup lacks a retry button and offers no configuration control. Switching to a YAML workflow fixes both:

   • You get the standard Re-run jobs button like any other Actions workflow
   • Full control over languages, triggers, schedule, and query suites
   • Disabling the native setup is required after adding the YAML version: Settings → Advanced Security (under Security) → CodeQL analysis (under Code scanning and Tools) → Disable

   

   GitHub generates a starter workflow under Security → Code scanning → Add tool → CodeQL, which you can trim down and commit to .github/workflows/codeql.yml.

Issue actions