DescriptionPURANDHAR SAIRAM MANNIDI
2018-08-14 01:20:57 UTC
Description of problem:
Nove compute service does not correctly report PFs and ignore few PFs until nova-compute service is restarted
Version-Release number of selected component (if applicable):
Red Hat OpenStack Platform 10
How reproducible:
Intermittently
Steps to Reproduce:
1. Spawn Multiple instances in the compute node with combination of PF + VF per instance.
2. restart nova-compute, delete instances
3. few PFs are not considered and doesn't report the availability during periodic resource audit.
Actual results:
Intermittently Nova fails to consider few PFs and it's corresponding VFs.
Expected results:
Irrespective of nova-compute restarts, nova-compute should consider the PFs in pci_passthrough_whitelist
Additional info:
It's a know issue. When PFs are attached to guest they are not listed when restarting nova-compute, the workaround is to restart the service.
There is a summary of the issue here; please can you confirm the behavior?
https://bugzilla.redhat.com/show_bug.cgi?id=1589796#c31
Thanks,
s.