[DRBD-user] Linstor: confining failure domains to a single back-end device

[Daniel Dehennin]

Roland Kammerer <roland.kammerer at linbit.com> writes:

>> > Do I understand correctly that it means, on each server:
>> >
>> >     1 disk == 1 PV == 1 VG == 1 ThinPool == 1 storage-pool
>> >
>> > ?
>> 
>> Ok, so I don't understand how to create and use this setup. Does someone
>> have any hints?
>
> I'd say your assumption is correct. The "handle" to create actual
> resources in LINSTOR then is the resource group/volume group (LINSTOR
> rg/vg, not to confuse with LVM). And LINSTOR RGs can have multiple
> LINSTOR storage pools. So you always use the same LINSTOR RG, it will
> choose one of the SPs and as there is a 1:1 mapping down to 1 physical
> disk your failure domain for that resource is that single disk.

Thanks a lot.

So, to see if I understood correctly:

1. when a first VM disk is created, one storage-pool is chosen to store it

2. any VM running, the disk will be “cloned” on the same storage-pool
   when possible and fallback to “copy” if the storage-pool is full[1]

3. Saving the running VM disk to the long term storage is always in
   “copy” mode[2], so no storage shared between them

Is that right?

Because I'm wondering how things are working when one storage-pool (==
one PV) is full.

Regards.

Footnotes:
[1]  https://github.com/OpenNebula/addon-linstor/blob/master/one/extender.py#L175-L182

[2]  https://github.com/OpenNebula/addon-linstor/blob/master/tm/cpds#L48-L53

-- 
Daniel Dehennin
Pôle Logiciels Libres du ministère de l’Éducation nationale
Récupérer ma clef GPG: gpg --recv-keys 0x5A380850F562870C
Empreinte: EEB2 C6C8 EDFE 8364 8B2B  A263 5A38 0850 F562 870C
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 442 bytes
Desc: not available
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20220823/e7f8b7d1/attachment.pgp>