[DRBD-user] DRBD8, dedicated repository for PVE 4

Jasmin J. jasmin at anw.at
Mon Dec 26 18:15:14 CET 2016

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.

Hello Dietmar!

 > The functionality you are looking for already exists using DRBD9. It is
 > called drbdmanage, and Linbit provides a repository including all
 > packages and a storage driver for PVE.
Yes it is ... BUT ...

 > AFAIK DRBD9 is stable (or will be soon?)
See here the post from Lars:
-> If you have a two node setup, stay with DRBD 8.4,
    you don't gain anything from the new features of DRBD 9,
    but, as you found out, still may be hit by the regressions.

Because of this post, I removed DRBD9 form my server, switched to DRBD 8.4
and wrote my Proxmox DRBD8 Storage Plugin. I would have loved to have no work
and to use the already existing tools.

 > that is simply not necessary (DRBD9 auto promotion ...)
I like that, but it is not available in DRBD 8.4. So a Proxmox DRBD8 Storage
Plugin would need to use dual primary mode and switch on both sides to primary.
When Proxmox PVE would execute on one server activate_volume and on the other
deactivate_volume a few seconds later, this would work. But then, if the
cluster falls apart and for what ever reason it gets activated on both sides
it might be a disaster. Therefore I simply disabled this in my Storage Plugin.
If someone likes to implement it, he can do this and test. My servers are now
productive and I have no reason to change a running system!
And ofcourse everybody is free to use DRBD9, drbdmanage and the Proxmox DRBD
Storge Plugin (for DRBD 9.x) provided by Linbit. In the long run THIS is the
preferred strategy.

 > ALL DRBD9 related code is now managed by Linbit. AFAIK they plan to provide
 > documentation soon on the DRBD site.
Yes, they did already announce it for Proxmox and the documentation about 9.x
is also very good (the 8.4 was also) and it is a very good thing. But as
written above, currently not 100% stable and recommended at the current state,
especially for a two node cluster.


More information about the drbd-user mailing list