runtime-config-flagd

Use the set_flag.py script with the flag key and variant, for example set_flag.py paymentFailure off. Always run with --dry-run first to preview the change, then apply it; flagd hot-reloads the ConfigMap within seconds.

Run list_scenarios.py --active-only to show only currently active failure-injection scenarios. The output includes the flag name, current variant, affected service, effect, and remediation steps, with a --json option for machine-readable output.

Available scenarios include payment failure and unreachability, ad service high CPU and GC pressure, email memory leak, image latency spikes, Kafka consumer lag, cache and catalog failures, cart errors, load generator traffic floods, and LLM inaccuracy or rate limiting.

Yes. When a service account token is mounted, the client builds a temporary kubeconfig file instead of passing the token on the command line, avoiding exposure via /proc. Outside the cluster it uses the default kubectl configuration.

Setting emailMemoryLeak to off stops further leaking, but already-allocated memory is not reclaimed. You must also restart the email pod to fully remediate the OOM condition.