Re: [pve-devel] [PATCH qemu-server v2 3/6] migration: fail when aliased volume is detected

[Fiona Ebner <f.ebner@proxmox.com>]

Am 24.05.23 um 16:40 schrieb Aaron Lauterer:
> On 5/22/23 16:17, Fiona Ebner wrote:
>> Am 12.05.23 um 14:40 schrieb Aaron Lauterer:
>>> Aliased volumes (referencing the same disk image multiple times) can
>>> lead to unexpected behavior in a migration.
>>
>> Not only migration, but snapshots, storage locking, etc. Should we
>> actually care here? I still think it is rather something that people
>> should be made aware for the storage layer. Maybe a big enough warning
>> in the documentation is enough?
>>
>> Since it's not only migration, should we add a warning during VM startup
>> instead/additionally?
>>
> 
> I guess a warning in the docs would be a low-hanging fruit -> added to
> my TODO.

Great!

> Snapshots should just fail the second time as we already have one with
> the same name, right?

I think so. But it's only when the same aliased disk is attached twice.

> AFAIU storage migration is a case where an aliased volume can lead to
> very unexpected behavior, like storages running full. So a check and die
> is probably a good idea.

That is a good argument for the check during migration :)

> An additional warning during startup could probably be a good idea. In
> that case, the checks should probably be factored out into their own
> method so we can call it from the start as well. Though that would
> probably mean, another round of iterating through the config instead of
> inlining it.

See my other reply to the original mail for 3/6 for a suggestion.