After upgrading Red Hat OpenShift, applications can misbehave due to several root causes: image or tag changes, workload shifts, instrumentation overhead, non-container-aware runtimes, cgroups v1-to-v2 migration failures, and unbounded deployments. The guide focuses on container awareness — how JVMs and native libraries like jemalloc or Netty detect (or fail to detect) container resource limits. Key scenarios include JVM failing to detect cgroups v2 limits after an upgrade to OpenShift 4.19 (which drops cgroups v1), and unbounded deployments where host CPU/memory directly inflates thread pools and heap sizing. Diagnostic steps include verifying cgroups compatibility, checking VM.info for heap and CPU detection data, and benchmarking to distinguish intermittent from persistent failures. A summary table maps each scenario to its diagnostic action and verification step.
