EKS Pod Operations

Read logs, find pods, exec into containers, and restart deployments across EKS environments.

When to use this skill

Use this skill when the user wants to:

• View logs from a pod or service
• List or find pods in an EKS environment
• Execute a command inside a running pod
• Restart (rollout restart) a deployment
• Debug a pod (describe, check status, view previous logs)
• Check if a service is running in a specific environment

Prerequisites

1. kubectl: installed and available in PATH
2. AWS CLI: installed with SSO profiles configured per environment
3. Kubeconfigs: generated per environment at ~/.kube/config_<env>
4. Config: ~/.eks-config.json — auto-generated on first run by scanning kubeconfigs and AWS profiles. See CONFIG.md for manual setup.

Pre-flight checks

Run the pre-flight script before the first operation in a conversation:

python3 <skill_dir>/scripts/eks_preflight.py

To also verify a specific environment (kubeconfig + SSO):

python3 <skill_dir>/scripts/eks_preflight.py --env stg

If SSO is expired, tell the user: aws sso login --sso-session <session> (session name is in the preflight output).

Operations

pods — List or describe pods

# Find pods for a service
python3 <skill_dir>/scripts/eks_ops.py pods --env stg --service my-service

# List all pods in environment
python3 <skill_dir>/scripts/eks_ops.py pods --env stg --all

# Describe a pod (events, conditions, mounts)
python3 <skill_dir>/scripts/eks_ops.py pods --env stg --service my-service --describe

logs — Read pod logs

# Last 100 lines (default)
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --service my-service

# Custom tail size and time window
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --service my-service --tail 500 --since 1h

# Previous container (after crash/restart)
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --service my-service --previous

# All replicas at once (uses stern if installed)
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --service my-service --all-pods

# Stream logs in real-time
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --service my-service --follow

# Specific pod by name (skip service resolution)
python3 <skill_dir>/scripts/eks_ops.py logs --env stg --pod my-service-abc123-xyz --tail 200

exec — Run a command inside a pod

Show the exact command to the user and get approval before running.

python3 <skill_dir>/scripts/eks_ops.py exec --env dev --service my-service -- ls /app/conf
python3 <skill_dir>/scripts/eks_ops.py exec --env dev --service my-service -- cat /app/conf/application.conf
python3 <skill_dir>/scripts/eks_ops.py exec --env stg --service my-service -- env | grep -i heap
python3 <skill_dir>/scripts/eks_ops.py exec --env stg --service my-service -- df -h

restart — Rollout restart a deployment

Show the restart command to the user and get approval before running.

python3 <skill_dir>/scripts/eks_ops.py restart --env dev --service my-service
python3 <skill_dir>/scripts/eks_ops.py restart --env dev --service my-service --watch

Important rules

1. exec and restart require explicit user approval. Show the exact command and wait for confirmation.
2. pods and logs are read-only. Run them directly without approval.
3. All output is automatically redacted — secrets, tokens, keys, and credentials are replaced with [REDACTED]. See references/REDACTION.md.
4. Bare env, printenv, or set commands are blocked inside exec. Use env | grep -i <keyword> instead.
5. Reading files under /secrets/, /vault/, or credential paths is blocked inside exec.
6. If [REDACTED] appears in output, do not attempt to retrieve the original value.
7. The script auto-selects the app container in multi-container pods (skips sidecars like statsite, istio-proxy).
8. Multiple replicas: when a service has multiple pods, --service picks the first running one. If the user wants a specific replica, run pods --env <env> --service <name> first to list all replicas, ask the user which one, then use --pod <pod-name> instead of --service.
9. Prefer Grafana or Splunk for production log investigation — use this skill as a complement.
10. Developers typically have read-only EKS access — restart may fail on production clusters.

Error handling

Troubleshooting