Area: Reliability
What: At the end of the integration suite, assert the OpenCost container restart count is zero; on failure, dump pod state and logs from the previous container instance.
Why: A panic or OOMKill can hide behind passing tests if the pod restarts quickly between requests. Restart count is a cheap, high-signal stability check across the whole run.
Acceptance criteria:
Restart count for OpenCost containers is 0 across the full suite run
On failure, output includes kubectl describe state and previous-container logs
Runs automatically as part of the standard suite
Related: restart recovery test (#12) — that one restarts deliberately; this one asserts none happen on their own
Area: Reliability
What: At the end of the integration suite, assert the OpenCost container restart count is zero; on failure, dump pod state and logs from the previous container instance.
Why: A panic or OOMKill can hide behind passing tests if the pod restarts quickly between requests. Restart count is a cheap, high-signal stability check across the whole run.
Acceptance criteria:
Restart count for OpenCost containers is 0 across the full suite run
On failure, output includes kubectl describe state and previous-container logs
Runs automatically as part of the standard suite
Related: restart recovery test (#12) — that one restarts deliberately; this one asserts none happen on their own