How do you do healthcheck to ApiSix deployment (liveness/readiness probes)

[bokjo]

Hello ApiSix people,

I'm trying to figure out how to do a health check on the ApiSix deployment itself.

Looking at the K8s/Helm implementation, the stateful set/pods do not have liveness/readiness probes set at all.

Is there a special port or endpoint on the GW(9080), AdminAPI (9180), or somewhere else for a simple /healthz check as others do?

I'm trying to deploy to GKE K8s through the new K8s Gateway implementation, but the health check always fails.

I've made a proxy on the GW(9080) /healthz to hosts like 127.0.0.1 and *.10 (since the GoogleHC sets the host to http://10.1.1.10 for example)... but no luck, ApiSix always says 404 route not found.

GW(9080) /healthz - just proxies the requests to Istio health check /healthz/ready port 15021

Any ideas are welcome!

p.s is there other way to get access to the Slack channel? send an email already while back, no luck... or it is meant only for @apache.org users?

[dosubot[bot]]

Hey @bokjo,

I'm here to help you with your ApiSix deployment question. I'm a bot that can assist you with bug fixes, answer questions, and guide you on how to become a contributor. Let me know how I can assist you!

Dosu

To perform a health check on an APISIX deployment using liveness/readiness probes, you can configure the probes in your Kubernetes deployment YAML file. The specific port and endpoint for the health check on the Gateway is port 9080 with the endpoint /healthz. Here is an example configuration:

apiVersion: v1
kind: Deployment
metadata:
  name: my-apisix-pod
spec:
  containers:
  - name: my-apisix-container
    image: my-apisix-image
    livenessProbe:
      httpGet:
        path: /healthz
        port: 9080
      initialDelaySeconds: 15
      periodSeconds: 10
    readinessProbe:
      httpGet:
        path: /readyz
        port: 9080
      initialDelaySeconds: 10
      periodSeconds: 5

This configuration sets up a liveness probe that checks the /healthz endpoint on port 9080 every 10 seconds to determine if the application is running. If the application is unhealthy, Kubernetes will restart the instance. The readiness probe checks the /readyz endpoint on the same port to determine if the application is ready to receive traffic ^[1].

Additionally, I found a similar closed issue where the suggested solution was to create a router /health by referring to the documentation at this link ^[2].

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other}

[boscard]

@bokjo have you found any solution? I'm trying to resolve the same problem...

[bokjo]

@bokjo have you found any solution? I'm trying to resolve the same problem...

@boscard I've created a route /healthz/ready in ApiSix that is essentially a proxy towards the Istio healthcheck in my case!

{
  "uri": "/healthz/ready",
  "name": "health-check",
  "desc": "Health check route for dev environment",
  "plugins": {
    "prometheus": {
      "_meta": {
        "disable": true
      },
      "prefer_name": false
    }
  },
  "upstream": {
    "nodes": [
      {
        "host": "istio-ingressgateway.***.svc.cluster.local",
        "port": 15021,
        "weight": 1
      }
    ],
    "type": "roundrobin",
    "hash_on": "vars",
    "scheme": "http",
    "pass_host": "pass"
  },
  "status": 1
}

Now the LB GKE HealthCheck is checking port 9080 on /healthz/ready

[boscard]

So if I understand correctly @bokjo , in your case health check status is returned not by ApiSix but by Istio?

[bokjo]

So if I understand correctly @bokjo , in your case health check status is returned not by ApiSix but by Istio?

Yes, but that makes sense in my case... If Istio is not healthy ApiSix/LB health check will fail and not serve traffic... also if ApiSix deployment is not up, the proxy HC will not go through, and traffic won't be served again.

If only ApiSix healthcheck is of your concern, you can make the /healthz/ready directly return a response with the response-rewrite plugin from ApiSix.