Issue in LessEqual
affecting enqueueability decision of proportion
plugin (in v1.4.0), requesting hotfix to v1.4.0
#2014
Labels
kind/bug
Categorizes issue or PR as related to a bug.
What happened:
When there are no jobs running in a queue, and the queue capability specifies a ScalarResource (i.e. gpu) attribute other than cpu/memory, the proportion plugin will always chose to reject enqueueing new jobs.
Problematic comparison code in
proportion
plugin:volcano/pkg/scheduler/plugins/proportion/proportion.go
Line 295 in 8423cb5
Bug (?) in
LessEqual
:volcano/pkg/scheduler/api/resource_info.go
Line 349 in 8423cb5
What you expected to happen:
For a queue with no jobs, the allocated and inqueue values were both 0 CPU, 0 Gi Mem.
Our queue specifies a capability of the form
nvidia.com/gpu is identified as a
ScalarResource
and appears to trigger the if condition here:volcano/pkg/scheduler/api/resource_info.go
Line 349 in 8423cb5
Thus any submitted jobs will be stuck in
Pending
state.How to reproduce it (as minimally and precisely as possible):
Try to schedule any job with the scheduler config below and a queue configured as specified above on a cluster with no running jobs.
Anything else we need to know?:
scheduler config
Environment:
kubectl version
):uname -a
):The text was updated successfully, but these errors were encountered: