Symptoms¶
- Multiple pods on
node-7.internalare being terminated unexpectedly. - Pod status shows
Evictedwith reasonThe node was low on resource: memory. - Application team reports intermittent 503 errors as pods are killed and rescheduled.
- The node has been flapping between
ReadyandMemoryPressureconditions over the past hour. - Monitoring shows node memory utilization spiking to 94-97% periodically.
- Evicted pods are being rescheduled to the same node, only to be evicted again.
- The cluster has 5 nodes, but resource requests are not set on many workloads.