CPU/MEM usage not accurate. Hits quotas while resources are available
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Juno |
Invalid
|
Undecided
|
Unassigned | ||
OpenStack Compute (nova) |
Won't Fix
|
Undecided
|
Unassigned | ||
Ubuntu |
Invalid
|
Undecided
|
Unassigned |
Bug Description
I tried to set my quotas to exactly the hardware specs of my compute node for accurate reporting. My compute node runs on bare metal. After I did this I got quotas exceeded, unable to launch instance. I checked the hypervisor from the web interface and cli. 5 of 12 VCPUs used and 6GB of 16GB used. When you try and change the quotas to match the hardware it errors. From the CLI it reports that quota cant be set to 12 VCPUs because 17 are in use. Same with mem says 17GB are in use. But they arent in use clearly. So the bandaid is to set the quotas really high and then it ignores them and allows you to create instances against the actual nova usage stats. This also causes some failure to launch instance errors, no host is available meaning there arent enough resources even though there are. Im running this as a production environment for my domain so I have spent hundreds of hours chasing my tail. Hope this is helpful for debugging the issue.
William, what version of OpenStack/Nova? Any specific distribution?