Skip to main content

Pod API latency

Last updated on

Pod API latency is a Kubernetes pod-level chaos fault that adds a configurable delay to selected API calls on the target pod for a configurable duration. It accepts a rich set of filters: HTTP path, method, headers, query parameters, source host/IP, destination host/IP/port, and direction (ingress or egress). It also supports HTTPS through user-supplied TLS certificates, so it can intercept and delay encrypted API calls. When the fault ends, latency returns to baseline immediately.

Use this fault when you need to slow down one specific dependency or endpoint while leaving the rest of the pod's traffic fast: simulate a slow upstream for payments.example.com, add latency only to POST /v2/users, or delay calls only from one tenant.

Run your first experiment

If you have not configured the chaos infrastructure yet, go to Quickstart to install the chaos infrastructure and run an experiment end to end.

Use cases​

Run this fault when you want to answer concrete questions like:

  • Timeout budget validation for one dependency: Add latency only to DESTINATION_HOSTS=payments.example.com to verify whether your code times out at the right moment and falls back without affecting other upstreams.
  • Path-specific tail latency: Slow down PATH_FILTER=/v2/checkout to test how a single hot endpoint's latency affects the calling page's overall load time.
  • Tenant-scoped degradation: Use HEADERS_FILTERS (for example X-Tenant: prod-eu) to inject latency for one tenant only.
  • Egress vs ingress impact: SERVICE_DIRECTION=egress slows the pod's outbound calls; SERVICE_DIRECTION=ingress slows requests coming into the pod.
  • HTTPS dependency testing: With TLS certificates supplied, you can add latency to encrypted gRPC and REST traffic the Pod HTTP latency fault cannot.

Prerequisites​

  • Kubernetes version: 1.21 or later. Go to What's supported to confirm distribution support.
  • Target pods are Running: The application pods you intend to target are in the Running state before the fault is launched.
  • Privileged pods allowed: The cluster lets you schedule privileged pods in the chaos namespace. GKE Autopilot supports this fault but requires the one-time setup in Chaos on GKE Autopilot; other locked-down distributions may need similar exemptions.
  • Container runtime access: The chaos pod can reach the container runtime socket on the target node (/run/containerd/containerd.sock, /var/run/docker.sock, or /var/run/crio/crio.sock).
  • API service on a known port: The target container serves HTTP, HTTPS, or gRPC traffic on a port you can specify with TARGET_SERVICE_PORT.
  • TLS material for HTTPS targets: When HTTPS_ENABLED=true, you provide CA, server, and (optionally) client certificate secrets so the proxy can terminate TLS and apply filters.
  • Workload selector defined: The chaos experiment knows the target workload by kind, namespace, and either names or labels.

Supported environments​

PlatformSupport status
Amazon EKSSupported
Azure AKSSupported
Google GKESupported
Red Hat OpenShiftSupported
RancherSupported
VMware TanzuSupported
Self-managed Kubernetes (CNCF-certified)Supported
GKE AutopilotSupported with Autopilot setup
EKS Fargate, ACI virtual nodesNot supported (no access to container runtime sockets)

Permissions required​

The fault runs under the chaos infrastructure's service account.

Resource (apiGroup)VerbsWhy it is needed
pods ("")get, list, create, delete, deletecollection, patch, updateDiscover target pods and run the chaos pod on the same node
pods/log ("")get, list, watchStream chaos pod logs for status and debugging
deployments, statefulsets, replicasets, daemonsets (apps)get, listResolve the target workload to the pods it owns
events ("")get, list, create, patch, updateRecord fault progress as Kubernetes events
jobs (batch)get, list, create, delete, deletecollectionRun the chaos job that drives the fault
secrets ("")get, listRead TLS certificate secrets when HTTPS_ENABLED=true

The default Harness chaos infrastructure service account already includes these permissions.

Fault tunables​

Configure the following fault parameters when you add Pod API latency to an experiment in Chaos Studio. Defaults are shown for reference.

Chaos parameters

TunableDescriptionDefault
LATENCYDelay to add to each matched API call. Accepts Go duration strings such as 200ms, 2s, 1m."2s"
TRANSACTION_PERCENTAGEPercentage of matched API calls to delay, between 0 and 100. 0 delays none; 100 delays every match.0
TARGET_SERVICE_PORTPort the target container listens on for API traffic.80
TOTAL_CHAOS_DURATIONDuration of the fault in seconds.60

Filters (API matching)

All filters are optional. Leave empty to match everything in that dimension. Combining filters narrows the match (AND across dimensions).

TunableDescriptionDefault
PATH_FILTERURL path the fault matches. Empty matches all paths.""
METHODSComma-separated HTTP methods to match (for example GET,POST). Empty matches all methods.""
QUERY_PARAMSQuery-parameter filter (for example tenant=prod). Empty matches all query strings.""
HEADERS_FILTERSHeader filter (for example X-Tenant: prod). Empty matches all headers.""

Filters (traffic source and destination)

TunableDescriptionDefault
SERVICE_DIRECTIONWhether to filter ingress traffic (received by the pod) or egress traffic (sent from the pod).ingress
SOURCE_HOSTSHostnames of the calling client (ingress only). Empty matches any source.""
SOURCE_IPSSource IPs of the calling client (ingress only). Empty matches any source.""
DESTINATION_HOSTSDestination hostnames for the call (egress only). Empty matches any destination.""
DESTINATION_IPSDestination IPs for the call (egress only). Empty matches any destination.""
DESTINATION_PORTSComma-separated destination ports (egress only). Empty matches any port.""

TLS (for HTTPS targets)

TunableDescriptionDefault
HTTPS_ENABLEDSet to true when the target serves HTTPS so the proxy terminates TLS to apply filters.false
CA_CERTIFICATESKubernetes secret holding the Base64-encoded CA certificate (ca.key, ca.crt) used by the proxy.""
SERVER_CERTIFICATESSecret holding the Base64-encoded server certificate (server.key, server.crt) the proxy presents.""
CLIENT_CERTIFICATESSecret holding the Base64-encoded client certificate (client.key, client.crt) for mTLS upstreams.""

Proxy and interface

TunableDescriptionDefault
PROXY_PORTPort the chaos proxy listens on inside the container's network namespace.20000
NETWORK_INTERFACENetwork interface inside the target container's namespace. Almost always eth0 for standard CNI plugins.eth0

Targeting

TunableDescriptionDefault
TARGET_PODSComma-separated list of pod names to target. Empty selects from the workload's pods using POD_AFFECTED_PERCENTAGE.""
TARGET_CONTAINERContainer in the pod whose network namespace to enter. Empty targets the first container in the pod spec.""
NODE_LABELLabel selector to filter target pods by the node they run on. Empty disables node-based filtering.""
POD_AFFECTED_PERCENTAGEPercentage of the workload's pods to target. 0 means one pod.0
SEQUENCEWhen multiple pods are targeted, inject parallel (all at once) or serial (one after another).parallel

Runtime and helper

TunableDescriptionDefault
CONTAINER_RUNTIMEContainer runtime on the target nodes. One of containerd, docker, crio.containerd
SOCKET_PATHPath to the container runtime socket on the target node. Set to match CONTAINER_RUNTIME./run/containerd/containerd.sock
RAMP_TIMEWait period in seconds before and after the fault. Go to ramp time to read how it is applied.0

Tunables that apply to every chaos fault are documented in common tunables for all faults.

Set LATENCY just past your client timeout

Picking a LATENCY value slightly larger than the caller's configured timeout (for example LATENCY=6s when the caller times out at 5s) reliably triggers the timeout path without making the experiment unnecessarily long.

Configure for your container runtime​

Set CONTAINER_RUNTIME and SOCKET_PATH to match the runtime on the target node:

CONTAINER_RUNTIMESOCKET_PATH
containerd (default)/run/containerd/containerd.sock
docker/var/run/docker.sock
crio/var/run/crio/crio.sock

Fault execution in brief​

Intercepts API traffic on TARGET_SERVICE_PORT inside the container's network namespace and adds LATENCY to the configured percentage of calls that match the path, method, header, query, and source or destination filters, optionally terminating TLS to apply the same logic to HTTPS calls.

Expected behavior during fault execution​

  • API calls that match every configured filter are delayed by LATENCY. Calls outside the filter set pass through with normal latency.
  • Only the percentage of matched calls selected by TRANSACTION_PERCENTAGE is delayed; the rest pass through unchanged.
  • For HTTPS targets, the proxy terminates TLS using the supplied certificates so it can read the request before delaying.
  • Clients with timeouts shorter than LATENCY see request timeouts (and may retry); clients with longer timeouts succeed but with elevated tail latency.
  • gRPC unary calls are delayed the same way HTTP responses are. Streaming RPCs are delayed at the next frame boundary.
When the fault ends

After TOTAL_CHAOS_DURATION, the proxy is torn down and API latency returns to baseline within a couple of seconds. Calls already buffered for delay are released as soon as cleanup runs.

Signals to watch​

Attach resilience probes to assert each layer:

  • Tail latency on the scoped path: Use a Prometheus probe on the API's p99 latency metric, filtered to the path you delayed, to confirm the increase matches LATENCY.
  • Client timeouts and retries: Use an HTTP probe against the calling service to detect timeout-induced errors.
  • Unrelated traffic stays fast: Use a Prometheus probe on the same p99 metric scoped to an unaffected path to confirm no collateral latency.

Verify the fault execution effect​

While the experiment is running, confirm the right calls are being delayed:

  1. Time a matched call and an unmatched call.

    kubectl run -n <namespace> tester --image=curlimages/curl --rm -it -- \
      curl -w "time=%{time_total}\n" -o /dev/null -s \
        "http://<target-pod-ip>:<TARGET_SERVICE_PORT><PATH_FILTER>"

    time_total should reflect the added LATENCY for matched calls. A request to a non-matched path should return at baseline latency.

  2. Confirm the proxy is intercepting the right port.

    kubectl exec -n <namespace> <target-pod> -- ss -tlnp

    The proxy listens on PROXY_PORT and reroutes TARGET_SERVICE_PORT through it for the fault's duration.

Recovery and cleanup​

  • End of duration: The proxy is removed automatically and API latency returns to baseline.
  • Abort the experiment: Stopping the experiment from Chaos Studio triggers the same cleanup path.
  • Failed cleanup: If automated cleanup did not complete, restart the target pod to reset its network state.

Limitations​

  • Serverless Kubernetes (EKS Fargate, ACI virtual nodes): These platforms do not expose container runtime sockets and reject the privileged access the fault needs. GKE Autopilot is supported once the one-time setup in Chaos on GKE Autopilot is in place.
  • Windows containers: This fault is supported on Linux pods only.
  • HTTPS without certificates: When HTTPS_ENABLED=true, the proxy must terminate TLS. If the supplied certificates do not chain to one the client trusts, the client will refuse the connection before any filter is applied.
  • Long-lived streaming RPCs: Delay applies to new frames once the proxy intercepts them; frames already in flight when the fault starts may not be delayed.
  • Port already bound: If PROXY_PORT collides with a port the target container is already using, the fault fails to start. Pick a port number outside the application's range.

Troubleshooting​

Pod API latency experiment stays Pending or never starts in Harness Chaos Engineering

Inspect the chaos pods in the experiment namespace with kubectl describe pod -n <chaos-namespace>. The most common causes are taints on the target node that the chaos pods do not tolerate, insufficient resources, or a PodSecurity admission policy blocking privileged pods. Add the required tolerations to the experiment or run in a namespace with privileged Pod Security level.

No API latency observed during pod-api-latency

The most common causes are: TRANSACTION_PERCENTAGE is 0 (default) so no calls are delayed; filters are over-specified and match no real traffic (broaden PATH_FILTER and METHODS); HTTPS_ENABLED is false but the target serves HTTPS, so the proxy cannot read the request to evaluate filters; or the supplied TLS certificates do not chain correctly. Re-run with TRANSACTION_PERCENTAGE=100 and a broad single filter to confirm the path.

TLS handshake fails for pod-api-latency in Harness Chaos Engineering

Verify the secrets referenced by CA_CERTIFICATES, SERVER_CERTIFICATES, and CLIENT_CERTIFICATES exist in the chaos namespace and contain Base64-encoded key/crt pairs (ca.key/ca.crt, server.key/server.crt, client.key/client.crt). The server certificate must include the target pod's service name in its SAN list. If the client uses cert pinning, this fault cannot delay its traffic without disabling the pin.

API latency persists after pod-api-latency ends

Automated cleanup did not complete. Restart the target pod to reset its network state. If the issue recurs, capture the chaos pod logs from the experiment namespace before the next run and share them with Harness support.