Diagnose repeated container termination with conflicting memory signals.

Instruction: Explain what the evidence supports and what you would verify before changing limits.

Context:

A fictional Pod restarts during batch requests. The last termination reason is OOMKilled; exit code is 137. Its container memory limit is 512 MiB. A coarse dashboard shows 350 MiB, while a higher-frequency working-set series briefly reaches about 510 MiB. No node eviction event is shown in the supplied excerpt.

Updated

Official answer available

Read the opening below, then unlock the full answer and practical guidance.

OOMKilled and the limit make container memory pressure the leading hypothesis...

Upgrade to view official answer

Related Questions