Observations deploying current nightly version of Che

[l0rd]

Summary

I have installed Che next on OCP (cluster-bot) using chectl server:deploy --platform=openshift. Here are some observations.

[Fixed] Eclipse Che operator description is outdated

Eclipse Che operator description says:

This operator installs PostgreSQL, Keycloak, Plugin registry, Devfile registry and the Eclipse Che server, as well as configures all three services.”

Description needs to be reviewed.

[Fixed] There are still 2 operator channels

Using the "eclipse-che-preview" CatalogSource I can see 2 channels “next” and “next-all-namespaces”. Shouldn’t be one only?

[Dedicated issue] Backup/Restore API

I can see the Backup/Restore API still there. If it doesn’t work with DevWorkspaces we should remove them.

[Fixed] The version of the DevWorkspace operator installed

The installed version of the DevWorkspace is 0.10.0 (I was expecting next / 0.12.0)

[Explained] Operands Pods

What’s “eclipse-che-preview” and the pod named “664dfaf464a4b88c439632189072fd9fd1fc3e2752141ea530151d--1-4pt5g” (spec.container.image is quay.io/eclipse/eclipse-che-openshift-opm-bundles:7.42.0-401.next)?

[Can't reproduce] Other issues I have noticed deploying Che from OperatorHub UI

I tried to deploy Che next without chectl, using the OperatorHub UI. I had to add to manually add the catalog source first:

kubectl apply -f - <<EOF
apiVersion: operators.coreos.com/v1alpha1
kind: CatalogSource
metadata:
  name: eclipse-che-preview
  namespace: openshift-operators
spec:
  sourceType: grpc
  image: quay.io/eclipse/eclipse-che-openshift-opm-catalog:next
  updateStrategy:
    registryPoll:
      interval: 5m
EOF

Deploying through the UI I noticed that

Operands are deployed in openshift-operators namespace

I have created CheCluster CR in namespace eclipse-che (default one) but che-gateway, registries, postgres, che-server, che-gateway, che-dashboard are deployed in openshift-operators namespace. That’s a lot of noise in this shared namespace.

I am not able to access the dashboard (although route exist and the gateway logs are ok)

Relevant information

No response

[tolusha]

[FIXED] There are still 2 operator channels

Fixed by rebuilding a catalog source without bundles related to next-all-namespaces channel.

[CAN'T REPRODUCE] Other issues I have noticed deploying Che from OperatorHub UI

1. Create a custom catalog source

$ kubectl apply -f - <<EOF
> apiVersion: operators.coreos.com/v1alpha1
> kind: CatalogSource
> metadata:
>   name: eclipse-che-preview
>   namespace: openshift-operators
> spec:
>   sourceType: grpc
>   image: quay.io/eclipse/eclipse-che-openshift-opm-catalog:next
>   updateStrategy:
>     registryPoll:
>       interval: 5m
> EOF
catalogsource.operators.coreos.com/eclipse-che-preview created

2. Install Eclipse Che operator in openshift-operators namespace (from next channel)

3. Create eclipse-che namespace

$ oc create namespace eclipse-che
namespace/eclipse-che created

4. Create CheCluster custom resource in eclipse-che namespace

5. Eclipse Che is deployed in eclipse-che namespace and accessable (fix is required: fix: Restore tlsSupport field in CheCluster CR sample che-operator#1278)

[tolusha]

[FIXED] Eclipse Che operator description is outdated

eclipse-che/che-operator#1283

[tolusha]

[FIXED] The version of the DevWorkspace operator installed

che-incubator/chectl#1913

[tolusha]

Backup/Restore API

Will be tracked separately #21006

[tolusha]

[FIXED] The version of the DevWorkspace operator installed

che-incubator/chectl#1913

[tolusha]

[Explained] Operands Pods

What’s “eclipse-che-preview” and the pod named “664dfaf464a4b88c439632189072fd9fd1fc3e2752141ea530151d--1-4pt5g” (spec.container.image is quay.io/eclipse/eclipse-che-openshift-opm-bundles:7.42.0-401.next)?

It is custom catalog source to deploy Eclipse Che from the next channel