While there are optimizations we can make to the test suite to mitigate this, sometimes resource contention causes things to slow to a crawl for reasons outside our control. In the example above, it spent 9m15s on "Free up unused disk space", which normally takes ~12s. The only thing that step is doing is running rm -rf, so that indicates extremely high I/O contention on the runner.
Modifications
This removes the timeout so it uses the default of 360 minutes (docs). This is what ArgoCD does (source), and E2E tests for ArgoCD PRs often take ~50 minutes (example).
Partial fix for #10807
Motivation
Occasionally, tests are being canceled due to exceeding the 30 minute timeout. Example: https://github.com/argoproj/argo-workflows/actions/runs/11358993638/job/31594490093
While there are optimizations we can make to the test suite to mitigate this, sometimes resource contention causes things to slow to a crawl for reasons outside our control. In the example above, it spent 9m15s on "Free up unused disk space", which normally takes ~12s. The only thing that step is doing is running
rm -rf, so that indicates extremely high I/O contention on the runner.Modifications
This removes the timeout so it uses the default of 360 minutes (docs). This is what ArgoCD does (source), and E2E tests for ArgoCD PRs often take ~50 minutes (example).
Verification
Wait for actions to run