Kubernetes Pod Management

Overview

Inspect pod status, health, events, and resource usage across cluster environments.

Pod Status

List pods for a service

kubectl --context=<env> get pods -n <namespace> | grep <service>

List all pods in a namespace with status

kubectl --context=<env> get pods -n <namespace> -o wide

Check for unhealthy pods

kubectl --context=<env> get pods -n <namespace> | grep -v "Running\|Completed"

Diagnostics

Pod details and events

kubectl --context=<env> describe pod <pod> -n <namespace> | tail -30

Events show recent scheduling, pulling, starting, and crash info.

Check restart counts

kubectl --context=<env> get pods -n <namespace> -o custom-columns=\
NAME:.metadata.name,STATUS:.status.phase,RESTARTS:.status.containerStatuses[0].restartCount,\
AGE:.metadata.creationTimestamp | grep <service>

Pod resource usage

kubectl --context=<env> top pods -n <namespace> | grep <service>

Environment variables (non-secret)

kubectl --context=<env> exec -n <namespace> <pod> -- env | grep -iv "pass\|secret\|token\|key" | sort

Check which image/version is running

kubectl --context=<env> get pod <pod> -n <namespace> \
  -o jsonpath='{.spec.containers[0].image}'

The image tag is the GitHub Actions run ID (databaseId). To find which PR/commit it corresponds to:

# Cross-reference image tag with CI runs on main
gh run list --repo <org>/<repo> --branch main --json databaseId,displayTitle,headSha \
  | jq '.[] | select(.databaseId == <image-tag>)'

Common Namespaces

Read namespace names from the project's CLAUDE.md (## Infrastructure section). Typical patterns:

• Application services: one namespace per cluster (e.g. plataformadato)
• Shared infrastructure (RabbitMQ, monitoring): separate namespace (e.g. shared-infra)

Rollout Management

Check deployment status

kubectl --context=<env> get deploy -n <namespace> | grep <service>

Restart a deployment (rolling)

⚠️ ALWAYS ask for explicit confirmation before running this command, in any environment.
State clearly: what deployment will restart, in which environment, and what the impact is (brief downtime during rollout).
Do NOT run without the user saying "yes", "proceed", or equivalent.

kubectl --context=<env> rollout restart deployment/<deployment> -n <namespace>

Check rollout history

kubectl --context=<env> rollout history deployment/<deployment> -n <namespace>