Containers, Kubernetes & Platform Engineering

A route-specific 404 may come from host or path matching, ingress class, service port, or application routing. Checks to make Compare ingress rules with service selectors and port names, inspect controller events and access logs, and test the service directly from within the cluster. Check exact hostname and path rewrite behavior before changing...

In-cluster DNS failures can result from CoreDNS health, search domains, network policy, or incorrect service naming. Checks to make Test resolution from the affected pod, check namespace and fully qualified service name, inspect CoreDNS pods and logs, and review network policy. Compare a known-good workload and avoid changing cluster DNS globally...

An image built for one CPU architecture may not run on a different node pool or may lack a matching manifest. Checks to make Check node architecture, image manifest platforms, scheduling constraints, and container runtime events. Build and publish required architectures or constrain workloads to compatible nodes, then verify rollout across intended...

CPU requests guide scheduling but do not cap usage, so workloads can consume spare node capacity or hit limits. Checks to make Compare usage with requests and limits, inspect throttling and node pressure, and identify workload or traffic changes. Set requests based on representative demand and use limits only with awareness of latency and throttling...

Node upgrades can expose runtime incompatibilities, changed kernel behavior, drained workloads, or insufficient disruption capacity. Checks to make Compare pod events and logs across upgraded and untouched nodes, verify runtime and kernel versions, review disruption budgets and resource availability, and check persistent volume attachment. Roll forward...

Mount failures can result from claim binding, storage class, access mode, node availability, or provider attachment limits. Checks to make Inspect pod events, PVC and PV state, storage class parameters, and node attachment conditions. Confirm the requested access mode and zone compatibility, and avoid deleting claims or volumes before data ownership is...

Slow image pulls can delay releases because of large layers, registry limits, network throughput, or node cache behavior. Checks to make Measure pull time by image and node, review image size and layer reuse, check registry throttling and network errors, and compare with cached pulls. Optimize images and registry access without bypassing image...

The problem may be service discovery, network policy, credentials, TLS, or database capacity. Checks to make Test DNS and port access from the affected pod, inspect network policies and service endpoints, and review database authentication and connection limits separately. Use a test identity and avoid placing credentials in diagnostic commands. When...