Unable to create workspace from sample on K8s due wrong URLs reported by devfile registry

[sleshchenko]

Describe the bug

Unable to create workspace from sample on K8s due wrong URLs reported by devfile registry:

^ see the cluster internal IP is reported.

Che version

• latest
• nightly
• other: please specify

Steps to reproduce

1. Deploy Che on K8s with helm chart (chectl server:deploy --installer=helm --platform=minikube --multiuser)
2. Open dashboard and check if samples have icons displayed.
3. Try to create a workspace with sample.

Expected behavior

Icons are displayed and it's possible to create a workspace with sample.

Runtime

• kubernetes (include output of kubectl version)
• Openshift (include output of oc version)
• minikube (include output of minikube version and kubectl version)
• minishift (include output of minishift version and oc version)
• docker-desktop + K8S (include output of docker version and kubectl version)
• other: (please specify)

Screenshots

see avove

Installation method

• chectl
  • provide a full command that was used to deploy Eclipse Che (including the output)
  • provide an output of chectl version command
• OperatorHub
• I don't know

Environment

• my computer
  • Windows
  • Linux
  • macOS
• Cloud
  • Amazon
  • Azure
  • GCE
  • other (please specify)
• Dev Sandbox (workspaces.openshift.com)
• other: please specify

Eclipse Che Logs

Additional context

[svor]

I don't see such problem with --installer=operator, only reproducible if use helm. @tolusha @mmorhun any idea what could be wrong?

[tolusha]

I see that che-server returns correct settings:

{
  "supportedRecipeTypes": "kubernetes,no-environment",
  "che.workspace.auto_start": "true",
  "cheWorkspacePluginRegistryUrl": "https://plugin-registry-eclipse-che.192.168.99.100.nip.io/v3",
  "cheWorkspacePluginRegistryInternalUrl": "http://plugin-registry.eclipse-che.svc:8080/v3",
  "cheWorkspaceDevfileRegistryUrl": "https://devfile-registry-eclipse-che.192.168.99.100.nip.io",
  "cheWorkspaceDevfileRegistryInternalUrl": "http://devfile-registry.eclipse-che.svc:8080",
  "che.factory.default_plugins": "redhat/vscode-commons/latest",
  "che.factory.default_editor": "eclipse/che-theia/next",
  "che.workspace.storage.available_types": "persistent,ephemeral,async",
  "che.workspace.storage.preferred_type": "persistent",
  "che.devworkspaces.enabled": "false"
}

Is that dashboard who adjust url ?

[svor]

@tolusha probably che-devfile-registry configmap should contain env variable with URL like here

[vitaliy-guliy]

linking with #19084

[tolusha]

@sleshchenko @svor
Everything was working in 7.32.0
The problem can't be with helm templates.
eclipse-che/che-server#48 is not relevant.

[svor]

@vitaliy-guliy Could you please check if the problem can be related to these changes eclipse-che/che-devfile-registry#420

[vitaliy-guliy]

@vitaliy-guliy Could you please check if the problem can be related to these changes eclipse-che/che-devfile-registry#420

Yes, checking it..

[vitaliy-guliy]

@sleshchenko @svor
Everything was working in 7.32.0
The problem can't be with helm templates.
eclipse-che/che-server#48 is not relevant.

Links to the devfiles were wrong for single-host deployments.
We form links, based on CHE_DEVFILE_REGISTRY_URL and it must be set into devfile registry container.

When deploying Che with helm, the variable is not set.

You could also open the conrainer logs and scroll to top.
There is a message, indicating missing variable.

WARNING: environment variable 'CHE_DEVFILE_REGISTRY_URL' not configured for an offline build of this registry. This may cause issues with importing projects in a workspace.

[tolusha]

I've just deployed 7.32.0
I see the very same message

WARNING: environment variable 'CHE_DEVFILE_REGISTRY_URL' not configured for an offline build of this registry. This may cause issues with importing projects in a workspace.

But I can create a workspace.

[svor]

It means that CHE_DEVFILE_REGISTRY_URL environment variable wasn't set for che-devfile-registry pod, probably the message should be improved. For my understanding CHE_DEVFILE_REGISTRY_URL should be defined in any case. If to compare configmap deployed via operator and helm we can see that env was set in operator deployment:

but it is absent in the helm deplpoyment (also in 7.32.0 version):

[tolusha]

@svor @vitaliy-guliy
We've decided to revert changes in devfile registry not to take into account CHE_DEVFILE_REGISTRY_URL if it isn't set into a container and not to update helmcharts (basically it keeps 7.32.x version behavior)