anuraagnalluri commented on pull request #369:
URL: 
https://github.com/apache/incubator-yunikorn-k8shim/pull/369#issuecomment-1073569993


   @yangwwei Apologies for the relative inactivity on this PR, just getting 
back to it now. Addressed most of your concerns and am going with the singleton 
approach since I could not find a way to share context between different ginkgo 
test suites (only within a suite via `Describe` block). Since ginkgo also has 
the ability to parallelize within a suite, `It` blocks potentially run within 
their own "containers". Therefore, if port-forwarding is managed by the go 
runtime, it should be in the setup for all tests now that issue REST calls to 
the scheduler svc.
   
   If we can verify that all suites execute in independent runtimes (which 
seems plausible from the log output of the checks), we don't even need to 
follow singleton pattern since each runtime have its own port forwarder.
   
   With the changes I currently have, some checks intermittently fail due to 
not being able to get allocations post scheduler restart. Many times, all 
checks are passing, but there is not 100% pass rate when re-running them. I've 
printed some debug information which shows that we can see the sleep-pod in the 
dev namespace for both cases, but the allocations are empty in the failing 
cases. You can search for "appsInfo allocations are" in failing checks to 
verify this.
   
   Any thoughts on what the cause of empty allocations when querying 
`/ws/v1/apps` upon scheduler restart may be? Haven't been able to reproduce the 
error locally. 


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


Reply via email to