Kubernetes issues

Common Kubernetes issues and how to troubleshoot them

Kubernetes issues often look simple at first — a pod restarts, a service stops responding or a deployment stalls — but the underlying cause can sit in application code, configuration, scheduling, networking, storage, resource pressure or a recent production change. This guide gives DevOps, SRE and platform teams a practical starting point.

Most common Kubernetes issues

Production teams repeatedly encounter the same failure patterns. The most useful first step is recognizing the symptom and narrowing the investigation path.

Pods

CrashLoopBackOff

A container starts and repeatedly crashes. Check container logs, exit codes, application dependencies, configuration and recent deployments.

Images

ImagePullBackOff

Kubernetes cannot pull the image. Verify the image name and tag, registry access, imagePullSecrets, network reachability and registry availability.

Memory

OOMKilled

A container exceeds its memory limit and is terminated. Compare memory use with requests and limits, inspect heap growth and check for leaks or bursts.

Scheduling

Pod stuck Pending

The scheduler cannot place the workload. Look for insufficient CPU or memory, taints, affinity rules, quota, PVC binding or unavailable nodes.

Nodes

Node NotReady

A node is no longer healthy enough to run workloads. Inspect kubelet health, network connectivity, disk or memory pressure and underlying infrastructure.

Health checks

Readiness or liveness probe failures

Verify probe path, port, timeout and application startup behavior. Failed readiness removes traffic; failed liveness can trigger repeated restarts.

Networking

Service or DNS failures

Check Service selectors, endpoints, CoreDNS, NetworkPolicy, CNI health, ingress and load-balancer status when workloads are healthy but unreachable.

Storage

PVC Pending or volume mount failures

Inspect storage class, access mode, capacity, zone constraints, CSI events and attachment or mount errors.

Security

RBAC Forbidden errors

Validate ServiceAccount, Role or ClusterRole permissions and the corresponding bindings for the resource and namespace being accessed.

Scaling

HPA not scaling

Check metrics availability, target utilization, requests, min/max replicas and whether the application signal actually reflects load.

Deployments

Rollout stuck or failed

Inspect rollout status, ReplicaSets, events, image changes, readiness failures and whether a recent configuration or secret update broke startup.

Capacity

CPU, memory or disk pressure

High utilization can cause throttling, eviction, failed scheduling and latency. Review resource pressure together with service impact and capacity headroom.

How to troubleshoot Kubernetes issues systematically

1. Scope the impact. Identify which service, workload, namespace or customer path is affected.

2. Inspect workload state. Review pod status, restart counts, deployment availability and unhealthy containers.

3. Read events and logs. Kubernetes events frequently expose scheduling, image, probe, volume and node failures; application logs provide the next layer of context.

4. Check recent change. Compare the incident window with deployments, configuration changes, scaling changes and infrastructure events.

5. Verify resources and scheduling. Look for CPU, memory, disk, quota, taints, affinity and capacity constraints.

6. Follow the traffic path. Validate Service endpoints, DNS, ingress, gateways, load balancers and NetworkPolicy.

7. Check storage and dependencies. Inspect PVCs, CSI events, databases, queues, APIs and other downstream services.

8. Decide on the next action. Roll back, scale, restart, repair configuration, fail over or escalate with the evidence already collected.

Kubernetes issues are often connected, not isolated

A single underlying condition can produce several symptoms at once. For example, memory pressure may lead to OOMKilled containers, failed probes, restarts, latency and reduced service capacity. Treating each alert independently creates noise; troubleshooting should connect symptoms around the affected service and timeline.

For deeper incident handling, read the Kubernetes incident response guide. For proactive signals, see the Kubernetes monitoring guide.

How Titansama approaches Kubernetes issues

Titansama brings Kubernetes monitoring, AI-assisted incident intelligence, deployment context and cloud-cost awareness into one operating view. The objective is to help teams move faster from a Kubernetes symptom to the issue that deserves engineering attention while keeping production decisions under human control.

Turn Kubernetes issues into
clearer engineering decisions.

See how Titansama supports Kubernetes monitoring, troubleshooting, incident response and operational intelligence.