We are looking into the running quotas, but they seems pretty odd. All my users have negative values for there running quotas. Is that normal ?
Versions of the related components and OS (frontend, hypervisors, VMs):
5.8.0 (problem was already there with 5.6.2)
Steps to reproduce:
Have existing VM
Upgrade to 5.6.2 or any version with running quotas supported.
Current results:
Let pick one user. This user has actually 2 RUNNING VM (4GB + 2GB) and 1 UNDEPLOYED VM (1GB). Quotas show:
RUNNING VM : -1 / -
RUNNING CPU: -1 / -
RUNNING MEMORY: -2097152KB / -
The -1 is because that user has not running quotas set up. If you add some quotas to that users, you will see the current number of running vms. Check this for more information.
So I’ve set a limit of 2 running VM for this user. This user can start his 3rd VM, he has now warning but the VM stay in PENDING mode. The admin sees RUNNING VM: 0 / 2 is that the correct behavior ?
It’s like the counter is initialized without taking care of previously deployed VM. So if I stop a running VM, counter starts to go into negative values. If I restart this VM, counter goes up to 0. I don’t remember having a positive value in all my users.
I can see that the user has the running vms quota set to 2, but he currently has no running vms because of this <RUNNING_VMS_USED><![CDATA[0]]></RUNNING_VMS_USED>. So please could you check that the user has more than 2 vms in RUNNING state?
Also, we can do a simple test:
Terminate the vms for the user.
Then instantiate a vm using that user, wait until it’s in RUNNING state, repeat the step one more time.
Try to instantiate it one more time (this is the third time) what happens? Do you see a message or the vm can be instantiated?
About the counter it should be restarted automatically, so maybe the fsck did something wrong, I will be aware of this, if it’s happen again I will check.
It’s seems that old users, maybe those created before the Running quotas functionnality, have all this problem. Some of theme are at -30 running VMs. So it can’t be fixed just by terminating/recreating VMs. Do you think that fsck could fix that at some time ?