On 01/13/2014 08:51 PM, Per Jessen wrote:
Per Jessen wrote:
I'm setting up a pair of frontend nodes with
a drbd device. The drbd
will have LVM on top and the LVs will be shared out to other systems
(backends) via iSCSI. I'm very new to drbd, although I did play with
it some years ago. In order to make a node the primary, it seems I
need heartbeat (or something else) to manage it? Presumably I can
also add the lvm commands somehwere.
The drbd website recommends pacemaker, which says it is built on
corosync or heartbeat. My question is - this seems like quite a bit
of work for just making sure one node is the primary, is there no
I should add, the applications on the frontends have failover via
round-robin DNS, so no action needed there.
If you want your DRBD resources to automatically fail over from one node
to another, you want to use Pacemaker. If you're happy manually
stopping LVM, manually bringing up another node as primary etc. using
the CLI tools when you yourself notice a failure, you can probably do so
without a clustering tool like Pacemaker, but I haven't tried that myself.
To clarify heartbeat/pacemaker/corosync project status slightly:
- pacemaker was split off heartbeat v2 some years ago (pacemaker is the
bit that manages your cluster resources, and heartbeat became "just"
a messaging layer at that point).
- heartbeat is effectively dead (we don't include it on *SUSE anymore).
- corosync is now the messaging layer of choice.
So these days, think: corosync+pacemaker for doing HA on Linux. The
problem is Google has a very long memory for old projects ;)
Senior Clustering Engineer
To unsubscribe, e-mail: opensuse-ha+unsubscribe(a)opensuse.org
To contact the owner, e-mail: opensuse-ha+owner(a)opensuse.org