Che error when starting more than one Theia workspaces

[nikosnikolaidis]

Dear @goncalorolo,

I cant start more than one workspace at the same time.
When I try to open more than one it stacks in the loading screen, and if I go to Che I see the following error for the new workspace that I try to start.

[goncalorolo]

It seems that Che does not support duplicate names for containers even if they refer to different workspaces. Therefore, that problem arises when the two services share the same devfile template. If you test with different devfiles, you will see that you are able to run more than one workspace at a time.

@philipreimer, is there anything you can configure in order to prevent this?

[reimer-atb]

@goncalorolo can you share more information? How does the devfile look like? The screenshot mentions a service named db. I assume that is a container that should be started?

AFAIK, all your che workspaces run in the same kubernetes namespace, for example goncalo-rolo-che-XYZ. And service names within a kubernetes namespace must be unique.

[goncalorolo]

@goncalorolo can you share more information? How does the devfile look like? The screenshot mentions a service named db. I assume that is a container that should be started?

The devfile template that is originating the error is here.

[reimer-atb]

I think the problem are the static names in the endpoints definition:

    endpoints:
      - attributes:
          discoverable: 'true'
          public: 'false'
        name: db

I get the same error like you if I just try to start the pre-defined sample workspace "Java with Spring Boot and MySQL" more than once.

If I check the service names that are started in my kubernetes namespace, I see:

NAME                          TYPE        CLUSTER-IP       EXTERNAL-IP   PORT(S)                                  AGE
db                            ClusterIP   10.100.240.208   <none>        3306/TCP                                 12m
server9j7qhkch-tools          ClusterIP   10.100.212.198   <none>        5005/TCP,8080/TCP                        12m
serverdd2ao9lq-jwtproxy       ClusterIP   10.100.175.71    <none>        4400/TCP,4401/TCP,4402/TCP,4403/TCP      12m

Apparently all services that are started by che have some random part in their name (9j7qhkch, etc.) whereas the db service does not.

Starting the sample workspace a second time will try to start another service with name db, which is not allowed because it would result in the same dns name db.<namespace>.

Can you assign some random part to the name?

[reimer-atb]

@philipreimer, is there anything you can configure in order to prevent this?

@goncalorolo I don't see any setting that would allow users to start workspaces in different namespaces or anything like that.

If I change the endpoint name db to something else before starting the second workspace both workspaces start without problems.

It seems the only way is to adjust the devfile before starting the workspace.

The devfile already contains: workdir: '${CHE_PROJECTS_ROOT}/web-java-spring-petclinic'

And CHE_WORKSPACE_ID or CHE_WORKSPACE_NAME should also be available. Can you append or prepend one of those to the endpoint name?

[goncalorolo]

That could be a way to do it. However, notice that the endpoint name is used in other places in the devfile. Changing that in every occurrence can be complex.

Will check if we can adapt the devfiles in order to take these names from a variable, so we can easily propagate the changes.