DRA: add e2e_node tests #124608
Labels
kind/feature
Categorizes issue or PR as related to a new feature.
sig/node
Categorizes an issue or PR as relevant to SIG Node.
triage/accepted
Indicates an issue or PR is ready to be actively worked on.
What would you like to be added?
#124323 reminded us that test coverage of checkpointing and the interaction between kubelet and DRA drivers could be improved. We can do error injection and/or add delays by extending
kubernetes/test/e2e/dra/deploy.go
Lines 420 to 431 in 4946c1f
NewDriver
call in E2E tests, perhaps by extending theconfigureResources
call such that it both configures resources and the driver.Here are some test scenarios that may be relevant. In all cases the test must ensure that NodePreparedResources succeeds before running a pod and NodeUnprepareResources succeeds for all claims which were previously prepared (not called out explicitly below). All pods must stop running.
Retry NodePrepareResources
Retry NodeUnprepareResources
Retry NodePrepareResources after restart
As above, but now also restart kubelet before allowing NodePrepareResources to succeed.
Retry NodeUnprepareResources
Same for NodeUnprepareResources.
Partial success for NodePrepareResources
Partial success for NodeUnprepareResources
Same for NodeUnprepareResources.
Partial success for NodePrepareResources with restart
Restart kubelet while the NodePrepareResources call for driver B is running.
Partial success for NodeUnprepareResources
Restart kubelet while the NodeUnprepareResources call for driver B is running.
Pod deletion during NodePrepareResources
Pod deletion during NodeUnprepareResources
/cc @klueska
/assign @bart0sh
/sig node
Why is this needed?
Better test coverage.
The text was updated successfully, but these errors were encountered: