Memory leak ends in an OOM kill
Container memory grows steadily until the kernel kills it at its limit.
ImpactThe container restarts and its in-flight requests fail
Can it be seen coming?Yes
RupturaCovered and tested on the lab
How to detect it
Last termination reason is OOMKilled.
Metrics and signals to watch:
container_memory_working_set_byteskube_pod_container_resource_limitskube_pod_container_status_last_terminated_reason
Can you see it coming?
Yes Precursor: Sustained growth of memory toward the limit.
What it hits next
callers of the workload.
What to do
Raise the memory limit or fix the leak.
Sources
- kubernetes-mixin KubePodCrashLooping
- Kubernetes termination reason OOMKilled
Where Ruptura stands
Ruptura reports this failure as the saturation failure mode, and it is measured by a pre-registered test on the lab.
Related failures
- Memory growth accelerates toward the limit
- Sudden memory jump past the limit
- Container killed by OOM again and again
- Container crashes at start, again and again
- Failures that come back at a regular interval
See it coming instead of reading about it
Ruptura is an open-source operator for Kubernetes and OpenShift that predicts the failures in this catalog that can be predicted, detects the others, and sends them to Alertmanager. Watch an OOM crash predicted 21 minutes early · ★ GitHub
Ruptura is an open-source operator for Kubernetes and OpenShift that predicts the failures in this catalog that can be predicted, detects the others, and sends them to Alertmanager. Watch an OOM crash predicted 21 minutes early · ★ GitHub