[DRBD-user] Your thoughts on a DRBD -> LVM2 (not CLVM) stack.

James Vanns james.vanns at framestore-cfc.com
Thu Feb 15 13:00:23 CET 2007

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


OK, I have been running the following stack well for testing purposes (read from bottom up) on a two-node active/active cluster:

Linux
Xen
CLVM
/dev/drbd0
/dev/sda
Linux

Basically running alternate Xen domUs on each physical machine like so:

node0 | node 1
--------------
dom0  | dom0
domU0 | domU1
domU2 | domU3

Each domU having their own CLVM logical volumes essentially never stepping on each others toes (because only one instance is running on one physical node at a time).

So, DRBD takes care of the underlying block replication so, assuming I shut down everything cleanly, I can actually reverse the whole setup (as in domU1 on node0 and domU0 on node 1 etc.).

So my question is this: why do I need CLVM? Can I just use LVM2? Although the domUs all run on top of their own LV they do belong to the same volume group (which of course sits on top of drbd because it is the drbd device that is pvcreated) - is this why CLVM is needed? For the locking of volume group meta data?

I'm curious. Anyone got any experience, input etc?

Jim

-- 

James Vanns
Systems Programmer
Framestore CFC Ltd.




More information about the drbd-user mailing list