Relationships
#3524 Flaky: process_bundle_prober_test 'stopping a host that has already exited lists no children under its pid' fails under full-suite load
Opened by zack · 10/15/2026· Shipped 10/16/2026
integration/process_bundle_prober_test.ts:600, the test added for swamp-club#3494, failed once in verify-build under the full parallel suite (run d5d4c041-f40c-4f5d-9698-ce4722ce8e1f, commit e6e4503f1, which changes only unrelated integration tests). The assertion at line 636, that listed.includes(pid) is false, saw true: listChildren was called with the killed host's pid even though the test waited for the host to be reaped before prober.release(). The same suite passed on the previous commit (run 44131ba4). Likely cause: release() still holds the host's pid from before the kill and lists its children when the reap and release race under load, or the waitFor liveness check sees the pid as gone before the prober's own exit tracking does. Expected: once the host has exited, release() never lists children under its pid.
Shipped
Click a lifecycle step above to view its details.
Sign in to post a ripple.