Check before migration

[xlnt]

I just found a serious issue that needs to be addressed ASAP. And it should be quite easy to implement as well.
When migrating a VM between hosts using shared storage there really should be check/verification done by proxmox to make sure that the nodes actually can access and use the share storage.

I just migrated a VM from one host to another where the other host did not have access to the shared storage (in this case NFS) and that totally killed the running VM and then it is not possible to migrate it back to the working host unless you fix the access to the shared storage.

So what we need is a small check when the migrate button is pushed to verify which hosts it can be migrated to and only those should be available for selection in the drop down list.

 

[e100]

It is possible to mark a particular shared storage as being available only to particular nodes. I use this to mark my DRBD volumes as being accessable only on the two nodes where they exist.

Proxmox does check that the destination is marked as having access to the shared storage.

I agree that an additional check to prevent the problem you describe would be a good enhancement.

 

[tom]

you need to define on the storage definition if a storage is available for the host.

if its not available and you try to migrate a VM to this host, you will see an error like this:

"error: storage 'lvm-drbd' is not available on node 'hp2' (500)"

but this does not prevent to migrate if the storage is not accessible due to other reasons. we added this request to https://bugzilla.proxmox.com/show_bug.cgi?id=12

 

[xlnt]

In this case I have 2 nodes up and running. Both are configured to use the same NFS storage and at the time when it was set up they both could access the NFS storage. But for some reason node2 has lost connection to the NFS storage but it does not show in web ui. So when I then migrate vm from node1 to node2 the migration starts and finish quickly in error ofcourse with the above result, a dead VM.

So what I think is needed is a storage check, heartbeat to monitor an verify that the configured nodes actually have contact and can use the shared storage.

Simply put, a storage heartbeat for all storage, shared or local. Have the nodes check every 3 seconds or make it an option to set the the heartbeat interval in the web ui. And also ofcourse when trying to migrate a heartbeat check to verify that the selected host actually can access the storage before you press the migrate button.