ensure delete returns only when the delete operation is fully finished

[victornoel]

Hi,

If I'm calling delete on a job for example, it takes time for the operation to really finish, but the delete call returns early, even if I used the FOREGROUND DeletionPropagation (not sure it's really linked to the matter ^^).

It would make sense to add an option so that the call only returns when everything is deleted as expected. A bit like the --wait parameter of kubectl works in that context (see kubernetes/kubernetes#42594).

[victornoel]

Related to #1474

[victornoel]

Potential workaround:

client.batch().v1().jobs()
    .inNamespace(namespace)
    .withName(name)
    .waitUntilCondition(Objects::isNull, 10, TimeUnit.SECONDS);

[manusa]

There's a different behavior between what the REST API does when performing a deletion with the foreground propagation policy and what kubectl does.

When you perform a kubectl delete ... --cascade='foreground, the command waits until every dependent resource is deleted to complete the execution.

However, when you perform the same operation using the REST API (with foreground PropagationPolicy), the requests returns as soon as the resource has been modified and the deletionTimestamp has been added to the metadata. I'm not sure if the client should take care of performing this wait/blocking for you.

Your suggestion is exactly what you need to do if you really need to provide blocking until the resource was effectively deleted.

[victornoel]

@manusa I understand. It makes sense indeed, thank you for the explanation, I will continue with my solution then!