Inconsistent behavior on logs page for different k8s master versions

[floreks]

Issue details

I'm using local-up-cluster with newest kubernetes HEAD. After deploying some card with wrong containerImage pod stays in pending state with errors. Logs page show Internal error(500). On older cluster we can see normal log entry saying that pod is not started and is in state pending.

Environment

Dashboard version: HEAD
Kubernetes version: v.1.2.0-alpha-8 (Newest HEAD)

Steps to reproduce

1. Deploy app with not existing containerImage.
2. Wait for card to show error status.
3. Go to logs page.
4. Check same page on older kubernetes version.

Observed result

Inconsistent behavior for logs page between different kubernetes versions.

Comments

It's hard to say what is the expected behavior here. Pod is actually in pending state so the older version is correct to show this entry but there are some errors related to this pod and newer version actually displays internal error with event error message. For me newest behavior seems more convenient.

Here are some screens.

Newest k8s:

Older k8s:

What do you think? @bryk @maciaszczykm @cheld

[bryk]

To me there should be no 500 error. Backend should handle this case and either say that there are no logs or display message that is currently displayed.

[floreks]

I'll take a look at that.

[floreks]

Changing this https://github.com/kubernetes/dashboard/blob/master/src/app/backend/logs.go#L84
to return err.Error(), nil solves this issue and logs page is shown normally:

Messages of course will differ between version so we shouldn't check for that. I'll have to adjust test also. Is this solution ok? I was also thinking about other solution but that would require complete redesign of logs backend :)

[floreks]

@bryk what do you think?

[bryk]

Sounds good to me. Ship it :)