[
https://issues.apache.org/jira/browse/YUNIKORN-3093?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Peter Bacsko updated YUNIKORN-3093:
-----------------------------------
Description:
There is a race condition inside {{TestAssumePodError}}:
{noformat}
err = cluster.waitForApplicationStateInCore("app0001", partitionName,
"Completing")
assert.NilError(t, err)
app := cluster.getApplicationFromCore("app0001", partitionName)
assert.Equal(t, 0, len(app.GetAllRequests()), "asks were not removed
from the application")
assert.Equal(t, 0, len(app.GetAllAllocations()), "allocations were not
removed from the application")
{noformat}
When the application is transitioning to Completing state, the ask from the
application object is not immediately removed, it's a separate step that is
triggered inside the shim ({{Task.releaseAllocation()}}). So we might be too
quick to call {{app.GetAllRequests()}} and it's still 1 instead of 0. Solution:
poll {{app.GetAllRequests()}} until it reaches 0.
was:
There is a race condition inside {{TestAssumePodError}}:
{noformat}
err = cluster.waitForApplicationStateInCore("app0001", partitionName,
"Completing")
assert.NilError(t, err)
app := cluster.getApplicationFromCore("app0001", partitionName)
assert.Equal(t, 0, len(app.GetAllRequests()), "asks were not removed
from the application")
assert.Equal(t, 0, len(app.GetAllAllocations()), "allocations were not
removed from the application")
{noformat}
When the application is transitioning to Completing state, the ask from the
application object is not immediately removed, it's a separate step that is
triggered inside the shim ({{Task.releaseAllocation()}}). So we might be too
quick to call {{app.GetAllRequests()}} and it's still 1 instead of 0. Solution:
poll {{app.GetAllRequests()}} until it reaches 0.
> Flaky test TestAssumePodError
> -----------------------------
>
> Key: YUNIKORN-3093
> URL: https://issues.apache.org/jira/browse/YUNIKORN-3093
> Project: Apache YuniKorn
> Issue Type: Task
> Reporter: Peter Bacsko
> Assignee: Peter Bacsko
> Priority: Minor
>
> There is a race condition inside {{TestAssumePodError}}:
> {noformat}
> err = cluster.waitForApplicationStateInCore("app0001", partitionName,
> "Completing")
> assert.NilError(t, err)
> app := cluster.getApplicationFromCore("app0001", partitionName)
> assert.Equal(t, 0, len(app.GetAllRequests()), "asks were not removed
> from the application")
> assert.Equal(t, 0, len(app.GetAllAllocations()), "allocations were not
> removed from the application")
> {noformat}
> When the application is transitioning to Completing state, the ask from the
> application object is not immediately removed, it's a separate step that is
> triggered inside the shim ({{Task.releaseAllocation()}}). So we might be too
> quick to call {{app.GetAllRequests()}} and it's still 1 instead of 0.
> Solution: poll {{app.GetAllRequests()}} until it reaches 0.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@yunikorn.apache.org
For additional commands, e-mail: issues-h...@yunikorn.apache.org
