IL-0017 UK Ministry of Justice · 2023-07-21kubernetesaws
A user reported stuck deployments in the support channel. Six nodes could not hand out pod addresses. After adding fresh nodes and draining the affected ones, the team compared CNI settings with a test cluster and found that prefix delegation was not enabled on the live cluster. Enabling it fixed allocation. Time to resolution was about four and a half hours. The follow-up is a check that confirms IP allocation works.
New deployments stayed in ContainerCreating on six nodes.
Symptom ? Pods stuck in ContainerCreatingIP address exhaustion
Root cause ? Configuration change
A user report in the platform's support channel.
The report states no lessons beyond the review action.
IL-0019 A module upgrade shrank node disks and filled them with imagessame root cause · shared Node pool, AWS VPC CNI