Bug with hard disks in CEPH pool ?

[illunis]

Hi Everyone
We are having a three node cluster ProxMox / CEPH and till version 3.4-6 everything was fine

We recentrly upgraded to 3.4-9 and problems started.

As you can see from the screenshots i am trying to delete a hard disk from a vm168 with a vm-101-disk2 and i get an error that the VM 221 is locked !!!!

Looking at the hard disk of the VM 221 we see that it is called vm-101-disk1.

Obviously there is a mix going on that was not happening till last version.

Regards

illunis

 

[dietmar]

As you can see from the screenshots i am trying to delete a hard disk from a vm168 with a vm-101-disk2 and i get an error that the VM 221 is locked !!!!

Maybe there is a backup job running? If not, try to unlock manually:

# qm unlock 221

I am not sure why VM 168 has a reference to vm-101-disk2, which is obviously created for VM101. And why do you
get an error message about VM 221 when you work on VM 168 ....

 

[illunis]

Thanks for the reply

There is no backup job running.
The machines for different reasons were renamed a few months ago and this is the reason they are pointing to different hard disks (the originals)

The VM is not locked.

And like i said that was not a problem till version 3.4-6

Regards
illunis

 

[illunis]

Even MORE problems

with the new version 3.4-9 when u remove a hard disk it does not take it to the UNUSED hard disks but it REMOVES it immediately !!!!

Please have a look at these problems and sort them out as soon as possible.

Regards

illunis

 

[dietmar]

with the new version 3.4-9 when u remove a hard disk it does not take it to the UNUSED hard disks but it REMOVES it immediately !!!!

Again, your config is seriously wrong! vm-101-disk2 does not belong to VM160 (or VM221). That is why it is not added as unused disk.

 

[illunis]

Hi again
my configuration was working with no problems at all till version 3.4-6

I can move my VM anywhere i want to another cluster and rename it. So i can rename a VM from 100 to 200 and still use a hard disk called vm-disk-1...I fi decided to remove it it would move to UNUSED with no problems.

I have also just now tested your theory and it is WRONG...

If i remove a hard disk that has the same name with a VM still gets removed.

Please fix the problems or reproduce what i am saying instead of trying to prove me wrong. We are using Proxmox more than 5 years with no problems.

Version 3-4.-9 is buggy.

regards

illunis

 

[spirit]

seriously,

you shouldn't have different id for your disk and your vms.

For example, if you create a new vm with id=100, and you delete it, you'll delete all disks with id=100, including disks attached to vm200.

It's quite easy to rename rbd volumes with rbd command cli.

 

[illunis]

seriously,

you shouldn't have different id for your disk and your vms.

For example, if you create a new vm with id=100, and you delete it, you'll delete all disks with id=100, including disks attached to vm200.

It's quite easy to rename rbd volumes with rbd command cli.

I am not arguing that...actually this is what i am doing at the moment....all i am saying is that configuration was working with no problems at all till version 3.4-6....

I will report back if changing VMs to corresponding Hard Disks solves the problem.

Regards
illunis

 

[illunis]

Hi everyone
So, after changing hard disks to correspond to VM numbers some problems were fixed but some remain.

Let me explain a bit more

We have two 4x cluster servers using ProxMox and CEPH

Cluster 1 is used for VMs and Cluster 2 is used for the hard disks of the VMs. Cluster2 has a pool of 54 hard disks.

Till version 3.4-6 there was no problem at all.

After the upgrade to version 3.4-9 ALL VMs have a boot delay of around 20 seconds....I will say again that NOTHING else has changed at the servers or the network side except the upgrade of ProxMox.

Any help is appreciated.

Regards

illunis

 

[illunis]

Hello again,

and the bug report on version 3.4-9 continues.
As mentioned above there is a problem with a boot delay of the VMs...now thout the problem extends to backing up a VM that has Hard Disks on another cluster. Although the VM is working fine and fully responsive when we try to take backup (manual or scheduled) it starts and never finishes...0/0 mbs as u can at the screenshot.

An again....everything was working fine till version 3.4-6.

Regards
illunis