Status progressing when status.loadbalancer.ingress is non empty and does not have Hostname OR IP

[peterbosalliandercom]

argocd app get grafana
Name:               grafana
Server:             https://k8s-api.test.appx.cloud
Namespace:          test-ops
URL:                https://localhost:8080/applications/grafana
Repo:               https://github.com/helm/charts.git
Target:
Path:               stable/grafana
Helm Values:        http://github-raw-content-service.demo:8080/k8s-test-cluster-config?filepath=grafana/helm-config.yaml
Sync Policy:        <none>
Sync Status:        Synced to  (722b19a)
Health Status:      Progressing

GROUP                      KIND                NAMESPACE  NAME                        STATUS  HEALTH
extensions                 PodSecurityPolicy   test-ops   grafana                     Synced  Healthy
                           Secret              test-ops   grafana                     Synced  Healthy
                           ConfigMap           test-ops   grafana                     Synced  Healthy
                           ServiceAccount      test-ops   grafana                     Synced  Healthy
rbac.authorization.k8s.io  ClusterRole         test-ops   grafana-clusterrole         Synced  Healthy
rbac.authorization.k8s.io  ClusterRoleBinding  test-ops   grafana-clusterrolebinding  Synced  Healthy
rbac.authorization.k8s.io  Role                test-ops   grafana                     Synced  Healthy
rbac.authorization.k8s.io  RoleBinding         test-ops   grafana                     Synced  Healthy
                           Service             test-ops   grafana                     Synced  Healthy
apps                       Deployment          test-ops   grafana                     Synced  Healthy
extensions                 Ingress             test-ops   grafana                     Synced  Progressing

Similarly, the ingress for grafana is likely empty for status.loadBalancer.ingress, so we consider this not healthy.

See:
https://github.com/argoproj/argo-cd/blob/master/docs/health.md

In our situation this status.loadBalancer.ingress is

status:
loadBalancer:
ingress:
- {}
For us this is a valid state, we do not have a hostname or IP in that field maybe thats the reason?

We do not use a loadbalancer so it is empty. I think this problem needs to be fixed right?

[k8sedu]

ingress not update it. if u use traefik and version above 1.7, it has config.

[jessesuen]

For us this is a valid state. We do not use a loadbalancer so it is empty. I think this problem needs to be fixed right?

@peterbosalliandercom - which ingress controller are you running? Do you mind also providing the ingress object YAML? So far I know of the following ingress controllers which have issues with setting status. I would like to understand the logic & behavior of other problematic controllers as well.

• contour - not yet supported - (see: Address not set on Ingress resource projectcontour/contour#403)
• traefik - disabled by default, but can be configured (see: problem: ingress health check always be processing #968 (comment))

If your controller cannot be configured in a way that it updates status.loadBalancer.ingress, then I think the following workaround can be done, where Ingress objects are deployed using a placeholder value for status.loadBalancer.ingress:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  name: test-ingress
spec:
  rules:
  - http:
      paths:
      - path: /testpath
        backend:
          serviceName: test
          servicePort: 80
status:
  loadBalancer:
    ingress:
    - hostname: placeholder

The other point is that we will be implementing #911 (comment) which will allow users the ability to override our built-in health assessments, including the ingress health assessment.

BTW, the Ingress health assessment logic is in-line with what is being proposed with sig-apps: kubernetes-sigs/application#78

[peterbosalliandercom]

We use https://github.com/kubernetes/ingress-nginx

[peterbosalliandercom]

Hereby an example ingress object Yaml file:

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  generation: 1
  labels:
    app: grafana
    app.kubernetes.io/instance: grafana
    chart: grafana-1.12.0
    heritage: Tiller
    release: grafana
  name: grafana
  namespace: test-ops
spec:
  rules:
  - host: xxxx
    http:
      paths:
      - backend:
          serviceName: grafana
          servicePort: 80
        path: /
  tls:
  - hosts:
    - xxxx
    secretName: xxxx
status:
  loadBalancer:
    ingress:
    - {}

[lcostea]

---

Later edit: Please disregard this, it seems we have a problem with the kong ingress controller, it wasn't deployed correctly

---

The same thing happens for kong ingress
Here is the ingress yaml which remains in Progressing status

apiVersion: extensions/v1beta1
kind: Ingress
metadata:
  annotations:
    kubectl.kubernetes.io/last-applied-configuration: |
      {"apiVersion":"extensions/v1beta1","kind":"Ingress","metadata":{"annotations":{},"labels":{"app.kubernetes.io/instance":"po-xx"},"name":"po-xx-xxxx","namespace":"xxxx"},"spec":{"rules":[{"host":"*.dev.x.x.x","http":{"paths":[{"backend":{"serviceName":"po-xx-xxxx","servicePort":80},"path":"/api/v1/xxxxxxx[/]?$"}]}}]}}
  creationTimestamp: 2019-01-11T14:45:22Z
  generation: 1
  labels:
    app.kubernetes.io/instance: po-xx
  name: po-xx-xxxx
  namespace: xxxx
  resourceVersion: "11513131"
  selfLink: /apis/extensions/v1beta1/namespaces/xxxx/ingresses/po-xx-xxxx
  uid: 863e8068-15af-11e9-91f1-02fd02d95eb8
spec:
  rules:
  - host: '*.dev.x.x.x'
    http:
      paths:
      - backend:
          serviceName: po-xx-xxxx
          servicePort: 80
        path: /api/v1/xxxxxxx[/]?$
status:
  loadBalancer: {}

[peterbosalliandercom]

BTW, the Ingress health assessment logic is in-line with what is being proposed with sig-apps: kubernetes-sigs/application#78
<

According to above it should be non empty and thats exactly what it is. The list is non empty. Shouldn't it be correct that way?

[jessesuen]

According to above it should be non empty and thats exactly what it is. The list is non empty. Shouldn't it be correct that way?

I think you're right. I think we may have interpreted this incorrectly. I think the correct behavior should be:

Progressing:

status:
  loadBalancer: {}

Healthy:

status:
  loadBalancer:
    ingress:
    - {}

[peterbosalliandercom]

Thank you for the reponse.

[bnouvelbmll]

Was kubernetes-sigs/application#78 approved ? It would be nice to have the status "not continously progressing " when not using "load-balancers" it seems "load-balancers" in the concept of k8s (which are always assumed to come from cloud-provider) and ingress are slightly different concepts, and that one is possible without the other.

[jessesuen]

If your service type load balancer never gets an ip, you can write a custom ingress health check that always returns Healthy

…

On Mon, Sep 2, 2019 at 4:12 AM Bertrand Nouvel ***@***.***> wrote: Was kubernetes-sigs/application#78 <kubernetes-sigs/application#78> approved ? It would be nice to have the status "not continously progressing " when not using "load-balancers" it seems "load-balancers" in the concept of k8s (which are always assumed to come from cloud-provider) and ingress are slightly different concepts, and that one is possible without the other. — You are receiving this because you modified the open/close state. Reply to this email directly, view it on GitHub <#997?email_source=notifications&email_token=ADAW76LQB6NZ2MSP2CTI7HLQHTYJRA5CNFSM4GO324ZKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD5VQVEQ#issuecomment-527108754>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ADAW76L7DXE3JPI2TY5XXLDQHTYJRANCNFSM4GO324ZA> .

[AndreLouisCaron]

@lcostea I also have this issue with the Kong ingress controller. Do you happen to remember what was not deployed correctly?

[prateekkhera92]

@jessesuen - I wrote below to exlude my service, but its not working :/

    core/v1/Service:
        health.lua: |
          hs = {}
          hs.status = "Healthy"
          return hs

[robotjoosen]

Was running into the same things, but your path for the service is incorrect.

As described in the docs, you need to use group/kind. The group for Service is empty, so instead of core/v1/Service: you just need to do Service:.

Service:
  health.lua: |
    hs = {}
    hs.status = "Healthy"
    return hs

@jessesuen - I wrote below to exlude my service, but its not working :/

   core/v1/Service:
        health.lua: |
          hs = {}
          hs.status = "Healthy"
          return hs