K8s Namespace Troubleshooter

Step 2: Pending Pod Root Cause Analysis

• Identify pending pods:
  kubectl get pods -n knowledge-graph --field-selector=status.phase=Pending

• For each pod, run:
  kubectl describe pod <pod-name> -n knowledge-graph

• Parse Events section to determine cause:

  • Insufficient cpu/memory → Resource exhaustion
  • 0/xx nodes are available → Node selector/taint/affinity mismatch
  • persistentvolumeclaim "xxx" not found → PVC issue
  • image can't be pulled → Not applicable (usually ImagePullBackOff, not Pending)

• Recommend fixes accordingly.

Step 2.1: CrashLoopBackOff Pod Diagnosis

• Identify CrashLoopBackOff pods: kubectl get pods -n knowledge-graph | grep -E "CrashLoopBackOff|Error"

• For each CrashLoopBackOff pod, perform deep analysis:

  A. Check container termination reason: kubectl get pod <pod-name> -n knowledge-graph -o jsonpath='{.status.containerStatuses[*].lastState.terminated}'

  Common termination reasons:

  • OOMKilled (exitCode: 137) → Memory limit exceeded
  • Error (exitCode: 1) → Application error
  • Error (exitCode: 137) → SIGKILL (usually OOM or external kill)
  • Error (exitCode: 139) → SIGSEGV (segmentation fault)
  • Error (exitCode: 143) → SIGTERM (graceful shutdown failed)

  B. Check current and previous container logs: kubectl logs <pod-name> -n knowledge-graph --tail=100 kubectl logs <pod-name> -n knowledge-graph --previous --tail=100

  C. Check resource configuration: kubectl get pod <pod-name> -n knowledge-graph -o jsonpath='{.spec.containers[*].resources}' | jq .

  D. Check for common issues:

  Termination Reason	Root Cause	Remediation
  OOMKilled	Memory usage exceeds limit	Increase memory limits in deployment/pod spec
  Error (exitCode: 1)	Application startup failure	Check logs for application errors, config issues
  Error (exitCode: 137)	SIGKILL (external kill or OOM)	Check OOM events, increase memory or fix memory leak
  Error (exitCode: 139)	Segmentation fault	Debug application, check for incompatible libraries
  Error (exitCode: 143)	SIGTERM handling issue	Increase terminationGracePeriodSeconds, fix signal handling

  E. For OOMKilled specifically:

  • Check current resource requests vs limits ratio

  • Recommend increasing memory limit (typically 2x current limit)

  • Example patch command:

    kubectl patch deployment <deployment-name> -n knowledge-graph --type='json' -p='[
      {"op": "replace", "path": "/spec/template/spec/containers/0/resources/limits/memory", "value": "<new-limit>"},
      {"op": "replace", "path": "/spec/template/spec/containers/0/resources/requests/memory", "value": "<new-request>"}
    ]'
    

  F. Check pod restart count and frequency: kubectl get pod <pod-name> -n knowledge-graph -o jsonpath='{.status.containerStatuses[*].restartCount}'

  High restart counts indicate:

  • Persistent issues requiring investigation
  • Potential liveness probe misconfiguration
  • Resource constraint problems

Step 2.2: ImagePullBackOff Pod Diagnosis

• Identify ImagePullBackOff pods: kubectl get pods -n knowledge-graph | grep -E "ImagePullBackOff|ErrImagePull"

• For each ImagePullBackOff pod, check:

  A. Get image details: kubectl get pod <pod-name> -n knowledge-graph -o jsonpath='{.spec.containers[*].image}'

  B. Check events for specific error: kubectl describe pod <pod-name> -n knowledge-graph | grep -A5 "Events:"

  C. Common causes and remediation:

  Error Message	Root Cause	Remediation
  "repository does not exist"	Wrong image name/tag	Verify image name and tag exist in registry
  "unauthorized"	Missing/invalid credentials	Create/update imagePullSecrets
  "manifest unknown"	Tag doesn't exist	Verify tag exists, check for typos
  "connection refused"	Registry unreachable	Check network, firewall, registry status
  "x509: certificate"	TLS/SSL issues	Add CA cert or configure insecure registry

  D. For private registry issues:

  Check if imagePullSecrets is configured

  kubectl get pod <pod-name> -n knowledge-graph -o jsonpath='{.spec.imagePullSecrets}'

Step 3: Resource Allocation Efficiency Audit

Please analyze the resource allocation in the Kubernetes namespace knowledge-graph and answer the following questions:

1. Is the current allocation of CPU and memory reasonable?
2. Are there any cases of resource wastage or resource shortages?
3. Suggestions for optimizing resource allocation.

• Commands:

  Get resource requests and limits for all pods

  kubectl get pods -n knowledge-graph -o custom-columns=
  'NAME:.metadata.name,CPU_REQ:.spec.containers[].resources.requests.cpu,CPU_LIM:.spec.containers[].resources.limits.cpu,MEM_REQ:.spec.containers[].resources.requests.memory,MEM_LIM:.spec.containers[].resources.limits.memory'

  Check actual resource usage (requires metrics-server)

  kubectl top pods -n knowledge-graph

• Analyze for common issues:

  • High memory requests with low limits → Risk of OOMKilled
  • Very low requests compared to limits → May cause scheduling issues
  • No limits set → Unbounded resource usage risk
  • Requests > actual usage → Resource wastage

Step 4: Event Log Inspection

Please check the events in the Kubernetes namespace knowledge-graph and answer the following questions:

1. Are there any error or warning events?
2. What are the specific reasons for these events?
3. Possible solutions.

• Command:
  kubectl get events -n knowledge-graph --sort-by=.lastTimestamp

• Filter for Type = Warning or Error: kubectl get events -n knowledge-graph --field-selector type=Warning --sort-by=.lastTimestamp

• Group by Reason and correlate with affected objects

• Map common reasons to solutions:

  Event Reason	Description	Remediation
  ImagePullBackOff	Cannot pull container image	See Step 2.2 for detailed diagnosis
  CrashLoopBackOff	Container keeps crashing	See Step 2.1 for detailed diagnosis
  BackOff	Container restart backoff	Check logs, may be OOMKilled or app error
  OOMKilled	Out of memory	Increase memory limits
  FailedScheduling	Cannot schedule pod	Check resources, node selectors, taints
  FailedAttachVolume	Volume attach failed	Check PVC, StorageClass, node issues
  FailedMount	Volume mount failed	Check PVC bound, node access to storage
  Unhealthy	Probe failed	Check liveness/readiness probe config
  NodeNotReady	Node is not ready	Check node status, kubelet logs
  Evicted	Pod evicted	Check node resources, pod priority

Step 5: Comprehensive Report Generation

Please generate a comprehensive report for the Kubernetes namespace knowledge-graph that includes:

• Summarize findings from all previous steps.
• Provide prioritized action items for remediation.