[DRBD-user] LINBIT getting involved in Heartbeat
florian.haas at linbit.com
Tue Nov 17 09:19:58 CET 2009
On 11/17/2009 07:37 AM, Romain CHANU wrote:
> Regarding the announcement of LINBIT getting involved in Heartbeat (cf.
> you recommend the use of Pacemaker over OpenAIS / Corosync since OpenAIS
> / Corosync is still in development and apparently (cf. link above) has
> currently some issues.
Pacemaker is a cluster resource manager.
Heartbeat is a cluster messaging layer.
OpenAIS/Corosync is a cluster messaging layer.
The two cluster messaging layers in the stack are interchangeable. The
cluster resource manager is not.
Thus, "recommending Pacemaker over OpenAIS / Corosync" (i.e., saying
Pacemaker is better than OpenAIS/Corosync) would be nonsense. We
recommend Pacemaker, and we are now offering our customers and users a
choice of cluster messaging layers.
> Could you tell us more about that?
> I am a bit confused because the guys from Pacemaker recommend OpenAIS /
Yes, and as far as I can see that is due to two reasons:
1. Heartbeat, up to this point, had no one who stood up for the
responsibility of maintaining the existing code base.
2. Heartbeat has scalability issues with larger clusters (about 8 nodes
Issue #1 is now resolved. Issue #2 is irrelevant for 2-node clusters
(which, as per http://www.clusterlabs.org/wiki/UsagePoll, constitute the
vast majority of Pacemaker clusters out there).
In contrast, OpenAIS/Corosync has two issues which are likely to affect
specifically DRBD based 2-node clusters, and happen to be largely
irrelevant for larger or non DRBD based clusters:
- redundant rings are suffering from breakage;
- cluster messaging does not support unicast.
Plus, there is one significant issue which hurts all existing Heartbeat
2.1.x users, regardless of whether they use DRBD or not:
- there is no upgrade path from Heartbeat 2.1.x clusters to Pacemaker
There's a bunch of people out there currently stuck on Heartbeat 2.1.x
because, at least that's our impression, they are reluctant to switch
communication stacks (Heartbeat->OpenAIS/Corosync) and move to a more
advanced cluster resource manager (Pacemaker) at the same time. These
users are using outdated software and get no bugfixes (with the
exception of Novell's SLES 10 customers). A supported Heartbeat
messaging layer allows you to retain Heartbeat, while moving from the
built-in 2.1.x CRM to Pacemaker.
I hope this clarifies things.
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 260 bytes
Desc: OpenPGP digital signature
More information about the drbd-user