do not loop forever when message when Gateway API is not available

[joaquimrocha]

See #2504 (comment) for context.
Like other views, we need to account for whether this is resource is available.

[farodin91]

@joaquimrocha Do you any link how it is done in different locations?

[joaquimrocha]

The useList function, which we use for most listings, returns results as null till the fetching is done. After that, if there's an error, we display it (otherwise the results are shown). E.g.:

headlamp/frontend/src/components/common/Resource/ResourceTable.tsx

Line 148 in be9e66b

const { items, error, clusterErrors } = resourceClass.useList({ namespace: useNamespaces() });

headlamp/frontend/src/components/common/Resource/ResourceTable.tsx

Line 164 in be9e66b

errorMessage={resourceClass.getErrorMessage(error)}

[joaquimrocha]

@farodin91 Let me know if you are taking a look or whether you prefer us to do it. Cheers.

[farodin91]

I'll try checkout these ideas in the evening.

[joaquimrocha]

Hi @farodin91 . We would like to release this week with a fix for this. Let me know if you were able to take a look or whether we should take it. Sorry for the pressure, just trying to understand how we should move.

[farodin91]

@joaquimrocha Go for it. My last week were pretty full.

[joaquimrocha]

@sniok URGENT: Turns out this is an existing error in the useKubeObjectList logic and so it's currently affecting Headlamp when there's no endpoint for a certain resource. There are two problems:

1. We are not throwing an exception when no endpoint is available and instead we pretend that the list of items is just empty.
   

   headlamp/frontend/src/lib/k8s/api/v2/useKubeObjectList.ts

   Line 401 in 0fb7ee9

   endpoint

2. There's no single error being exported from useList, and thus this line in ResourceTable is wrong:
   

   headlamp/frontend/src/components/common/Resource/ResourceTable.tsx

   Line 148 in be9e66b

   const { items, error, clusterErrors } = resourceClass.useList({ namespace: useNamespaces() });

We should also probably fix the issue that we're not passing the endpoints for the desired cluster when querying.

[joaquimrocha]

Fixed with #2766