istio:mesh-selfheal
Skill304 starsChanged 4 months agoApache-2.0View on GitHub
Detect and recover from Istio Ambient mesh outages where ztunnel/waypoint proxies serve expired mTLS certs and every gateway route returns HTTP 503 (long-running or suspended dev/Kind clusters)
In this repository110 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- auth:keycloak-confidential-clientSkill
- auth:mlflow-oidc-authSkill
- auth:otel-oauth2-exporterSkill
- authSkill
- ci:monitoringSkill
- ciSkill
- ci:statusSkill
- cve:brainstormSkill
- cve:scanSkill
- cveSkill
- design:reviewSkill
- designSkill
- docs:reviewSkill
- genai:semantic-conventionsSkill
- genaiSkill
- git:commitSkill
- github:dependabotSkill
- github:issuesSkill
- github:last-week-orgSkill
- github:last-weekSkill
- github:my-statusSkill
- github:prsSkill
- githubSkill
- github:streamerSkill
- git:rebaseSkill
- gitSkill
- git:statusSkill
- git:worktreeSkill
- graph-loopSkill
- helm:debugSkill
- helmSkill
- hypershift:clusterSkill
- hypershift:debugSkill
- hypershift:preflightSkill
- hypershift:quotasSkill
- hypershift:setupSkill
- hypershiftSkill
- istio:ambient-waypointSkill
- istioSkill
- k8s:healthSkill
- k8s:live-debuggingSkill
- k8s:logsSkill
- k8s:podsSkill
- k8sSkill
- kind:clusterSkill
- kindSkill
- local:full-testSkill
- localSkill
- local:testingSkill
- metaSkill
- meta:write-docsSkill
- onboard:linkSkill
- onboardSkill
- onboard:standardsSkill
- openshift:debugSkill
- openshift:routesSkill
- openshiftSkill
- openshift:trusted-ca-bundleSkill
- orchestrate:ciSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Scope: dev / Kind: Do not auto-run recovery against a production mesh.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Istio Ambient Mesh Self-Heal
Detect and recover the classic "everything returns 503" outage on a long-running or
suspended dev/Kind cluster (rossoctl/rossoctl#1899): after a host suspend the Istio Ambient
data plane (ztunnel + waypoints) keeps serving expired istiod-issued mTLS certs and never
re-fetches, so all *.localtest.me routes 503 — while every pod still looks Running.
Scope: dev / Kind. Do not auto-run recovery against a production mesh.
When to use
Reach for this when:
http://*.localtest.me:8080/(UI, API, MLflow, Kiali…) all return HTTP 503, andkubectl get pods -Alooks healthy, Gateway/HTTPRoute status isAccepted/ResolvedRefs, and- the 503 body is
upstream connect error … connection terminationwith headerserver: istio-envoy.
That pattern = the data plane is reachable and routing, but the upstream mTLS hop fails on an expired cert.
Two failure domains (don't conflate)
- Mesh 503 (this skill). ztunnel/waypoint certs are istiod-issued (SDS), not SPIRE. The durable fix is upstream (istio/ztunnel#1679); the dev-cluster mitigation is to restart the data plane so it re-fetches fresh certs.
- SPIRE agent reattest deadlock (a separate, coincident suspend casualty) breaks
AuthBridge / token-exchange identity, not mesh routing. Check it with
--include-spire.
Detect + recover (one command)
# Detect only (read-only) — reports status and prints the recommended recovery commands
scripts/k8s/mesh-recover.sh
# Apply the recovery (rollout restart ztunnel + waypoints + gateway)
scripts/k8s/mesh-recover.sh --fix
# Also check/restart the SPIRE agent (separate AuthBridge domain)
scripts/k8s/mesh-recover.sh --fix --include-spireExit codes: 0 healthy · 2 degraded (or fix attempted) · 3 inconclusive (unreachable / not Ambient) · 1 usage/precondition error.
Useful flags: --probe-host <host>, --gateway-url <url> (default http://127.0.0.1:8080),
--json (machine-readable). The script discovers the ztunnel namespace (Kind installs put it
in istio-system; the chart uses istio-ztunnel) rather than hardcoding it.
Manual fallback (what --fix runs)
# 1. ztunnel — the node proxy all mesh traffic flows through (discover its namespace first)
ZT_NS=$(kubectl get ds -A -l app=ztunnel -o jsonpath='{.items[0].metadata.namespace}')
kubectl rollout restart daemonset/ztunnel -n "$ZT_NS"
# 2. waypoint proxies (L7), if any
kubectl rollout restart deploy -A -l gateway.networking.k8s.io/managed-by=istio.io-mesh-controller
# 3. the ingress gateway
kubectl rollout restart deploy/http-istio -n rossoctl-system
# verify
curl -s -o /dev/null -w "HTTP %{http_code}\n" -H "Host: rossoctl-ui.localtest.me" http://127.0.0.1:8080/Diagnostic one-liners
# Is it a mesh cert problem? (CertificateExpired / expired peer certificate)
kubectl logs -n "$ZT_NS" -l app=ztunnel --tail=100 | grep -iE "certificate expired|CertificateExpired"
# ztunnel cert status (if istioctl is installed)
istioctl ztunnel-config certificates "$(kubectl get pod -n "$ZT_NS" -l app=ztunnel -o jsonpath='{.items[0].metadata.name}')" -n "$ZT_NS"
# SPIRE agent failing to re-attest? (separate domain)
kubectl logs -n spire-system -l app=spire-agent --previous --tail=40 | grep -iE "reattest|token has expired|Agent crashed"Testing (dev/Kind)
A real suspend/clock-expiry can't be forced deterministically, so a helper induces the same
symptom (CA down + ztunnel restarted → data plane has no valid cert → mesh-wide 503) and runs the
full detect → --fix → re-probe loop. Destructive to the mesh; Kind-guarded.
scripts/k8s/induce-mesh-outage.sh --self-test # break → assert detect(exit 2) → restore CA → --fix → assert green
scripts/k8s/induce-mesh-outage.sh --break # induce and leave broken (to test manually)
scripts/k8s/induce-mesh-outage.sh --restore # undoReferences
- rossoctl/rossoctl#1899 — the mesh-wide 503 outage this addresses.
- istio/ztunnel#1679 — the upstream bug for the durable data-plane self-heal fix.
- Related skills:
k8s:health,istio:ambient-waypoint.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

