[pve-devel] avoiding VMID reuse

[Thomas Lamprecht]

Hi,

On 03/12/2018 03:13 PM, Lauri Tirkkonen wrote:
> Hi,
> 
> we'd like to keep VMIDs always unique per node, so that we can use zfs
> for backups more easily (ie. avoid dataset name collisions; granted,
> this still leaves room for collisions if disks are removed and added to
> a VM). We've implemented the following (crappy) PoC that accomplishes
> just this by storing the highest allocated VMID and returning that for
> /cluster/nextid.
> 

Sorry if I misunderstood you but VMIDs are already _guaranteed_ to be
unique cluster wide, so also unique per node?

Your approach, allowing different nodes from a cluster to alloc
the same VMID also should not work, our clustered configuration
file system (pmxcfs) does not allows this, i.e. no VMID.conf
file can exist multiple times in the qemu-server and lxc folders
()i.e., the folders holding CT/VM configuration files)

If your problem is that you request multiple VMIDs at the same
time, and thus run into a race, then take  a look at:

https://pve.proxmox.com/pipermail/pve-devel/2016-October/023440.html

I tried to address this when I tried to make the VMID on VM/CT
create optional. Making create optional won't happen, as it
violates the POST only once principle and brings you in hell's kitchen.

But the reserve VMID part may make some sense, at least it's idea.