Openshift connector and Che: the extension is not keeping the state after workspace restart.

[sunix]

Describe the bug

When using the openshift connector, the state is lost after restart.

Che version

• latest
• nightly
• other: please specify

Steps to reproduce

1. Start a workspace with the openshift connector plugin + a nodejs project
2. Login to a remote openshift cluster from the openshift connector panel
3. Create some project / component to the nodejs project, push
4. Restart the workspace
5. User is not connected anymore, and after reloaded, the component has no context although .odo folder is still present in the nodejs project.

Runtime

• Openshift 4.2

Installation method

• che-operator (operator hub)

[tsmaeder]

@sunix are you using persistent storage or ephemeral? Is the state preserved when you use the extension in VS Code?

[sunix]

Persistent storage. Yes it works on VS Code

[azatsarynnyy]

It's because of kubeconfig isn't persisted between workspace restarts. Should be fixed by #14877

[sunix]

@azatsarynnyy i don't see how these are related: #14877 is about chectl

[azatsarynnyy]

If I understand it right, the solution #2 , proposed in #14877, should fix that problem partly. When access to a cluster is set up with chectl workspace:inject.

[tsmaeder]

Note that any VS Code extension can save state to a directory provided by the API, so the only general fix is to preserve that directory between workspace runs.

[vinokurig]

Can not reproduce the bug. Looks like since we have the auto login feature, it is not relevant any more. Here is my steps to reproduce:

[sunix]

Hey sorry, @vinokurig autologin is for login in the current openshift cluster. What if we log in another cluster?

[nickboldt]

is this on track for Che 7.16? We'd like this to work OOTB in CRW 2.3.

[sunix]

also ... has autologin #15261 been moved to CRW?

[vinokurig]

@sunix

also ... has autologin #15261 been moved to CRW?

It should work for CRW.

[vinokurig]

Proposal: to keep the plugin state between workspace restarts we should have an additional PVC, like the /projects folder mount. @skabashnyuk WDYT?

[sunix]

@vinokurig yep so which folder(s) would you turn into volumes ?

[vinokurig]

It think it should be a new folder somewhere in root, don't know the name yet :)

[skabashnyuk]

keeping the state of workspace between restarts in a volume sounds good to me

[tsmaeder]

Note that we sometimes run multiple copies of VS Code extensions in various containers (VSCode YAML, for example). This could lead to the plugins overwriting each other's state. So if this is going to be a general fix, we should follow up by doing something about that: IMO, running multiple copies of the same plugin is wrong anyway: we needed it because we could not remotely call plugin API. Now that we can, we should refactor our plugins and prevent multiple copies of the same plugin.

[vinokurig]

The openshift-connector plugin uses the openshift-client tool under the hood, so the login state is kept in the config of the openshift-client tool, not in the plugin state. I've added a volume mount to the openshift-connector plugin which will save the config of the openshift-client tool, so the the login session will be saved between workspace restarts.