Can't apply managementApiAuth to existing datacenter

vsoloviov commented 10 months ago

What happened?

When I added

    managementApiAuth:
      manual:
        clientSecretName: cass-management-api-client
        serverSecretName: cass-management-api-server

to existing datacenter it failed to update sts because API is still running on http, but operator sends https requests:

2023-10-20T06:52:06.234Z    ERROR   controllers.CassandraDatacenter calculateReconciliationActions returned an error    {"cassandradatacenter": "cass-db/dc2", "requestNamespace": "cass-db", "requestName": "dc2", "loopID": "14a47b0d-9f66-45c4-a330-9ef440c8754a", "error": "Post \"https://10.1.3.11:8080/api/v0/ops/seeds/reload\": http: server gave HTTP response to HTTPS client"}
github.com/k8ssandra/cass-operator/internal/controllers/cassandra.(*CassandraDatacenterReconciler).Reconcile
    /workspace/internal/controllers/cassandra/cassandradatacenter_controller.go:149
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).Reconcile
    /go/pkg/mod/sigs.k8s.io/controller-runtime@v0.14.6/pkg/internal/controller/controller.go:122
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).reconcileHandler
    /go/pkg/mod/sigs.k8s.io/controller-runtime@v0.14.6/pkg/internal/controller/controller.go:323
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).processNextWorkItem
    /go/pkg/mod/sigs.k8s.io/controller-runtime@v0.14.6/pkg/internal/controller/controller.go:274
sigs.k8s.io/controller-runtime/pkg/internal/controller.(*Controller).Start.func2.2
    /go/pkg/mod/sigs.k8s.io/controller-runtime@v0.14.6/pkg/internal/controller/controller.go:235

What did you expect to happen?

I expected operator to communicate with pods and update configuration to enable mTLS on both operator and mgmt api sides.

How can we reproduce it (as minimally and precisely as possible)?

Bring up a basic datacenter with

    managementApiAuth:
      insecure: {}

Once datacenter is up and running create certificates chain

### MGMT API Auth
---
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: cass-management-api-ca
spec:
  isCA: true
  commonName: cass-management-api
  secretName: cass-management-api-ca
  duration: 87600h0m0s # 10y
  privateKey:
    encoding: PKCS8
  issuerRef:
    name: cassandra-ca-issuer
    kind: Issuer
    group: cert-manager.io
---
apiVersion: cert-manager.io/v1
kind: Issuer
metadata:
  name: cass-management-api-issuer
spec:
  ca:
    secretName: cass-management-api-ca
---
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: cass-management-api-server
spec:
  secretName: cass-management-api-server
  duration: 8760h0m0s # 1y
  issuerRef:
    name: cass-management-api-issuer
  dnsNames:
    - cass-management-api-server
  commonName: cass-management-api
  privateKey:
    encoding: PKCS8
---
apiVersion: cert-manager.io/v1
kind: Certificate
metadata:
  name: cass-management-api-client
spec:
  secretName: cass-management-api-client
  duration: 720h0m0s # 30d
  issuerRef:
    name: cass-management-api-issuer
  commonName: cass-management-api-user
  privateKey:
    encoding: PKCS8

Update datacenter configuration:

    managementApiAuth:
      manual:
        clientSecretName: cass-management-api-client
        serverSecretName: cass-management-api-server

cass-operator version

1.17.2

Kubernetes version

1.27.4

Method of installation

Argo

Anything else we need to know?

No response

┆Issue is synchronized with this Jira Story by Unito

vsoloviov commented 10 months ago

Similar issue happens if I build a datacenter with manual managementApiAuth and then disable it: it sends an http request to the https api endpoint

burmanm commented 9 months ago

Actions to take here:

Instead of checking CassandraDatacenter's status of the HTTPS, we need to check if ObservedGeneration < Generation of CassandraDatacenter and if the target Pod has HTTPS enabled or not and build the client security based on that information.
The above can be checked from running configs by inspecting the Pod and/or if the target server secret is mounted there. Can be found from Pod's ENV variables: MGMT_API_TLS_CERT_FILE is set (or a volume called "management-api-server-certs-volume" is mounted to the pod)

Miles-Garnsey commented 5 months ago

@burmanm is this ticket still current? It is from last year so I'm confirming before I start work on it.

Also, can you clarify what you mean by check if ObservedGeneration < Generation of CassandraDatacenter? What ObserverGeneration are we checking? Why would the ObservedGeneration and "Generation" be different if you're referring to the CassandraDatacenter in both cases?

burmanm commented 5 months ago

| @burmanm is this ticket still current? It is from last year so I'm confirming before I start work on it.

Yes, it is.

| Also, can you clarify what you mean by check if ObservedGeneration < Generation of CassandraDatacenter? What ObserverGeneration are we checking? Why would the ObservedGeneration and "Generation" be different if you're referring to the CassandraDatacenter in both cases?

Whenever you update the CassandraDatacenter, the ObservedGeneration is less than Generation until the reconciliation has finished. In this case, it wouldn't since it fails midway. However, I think you can just skip that check for now and simply verify if the target pod has https enabled or not for mgmt-api.

Currently, the syntax is func BuildManagementApiSecurityProvider(dc *api.CassandraDatacenter), yet this isn't sufficient. On the other hand, be careful that the SecurityProvider is the HTTPS one when constructing the PodTemplateSpec and other required modifications as enabling auth would require the https one to provide those curl commands and other properties. So it's not just the httphelper client.

k8ssandra / cass-operator