QUESTION: Cluster Stack!
February 22nd, 2010 — roaksoaxI was wondering how many people out there use loadbalancers based in Heartbeat/ldirectord/ipvsadm? If you do in Ubuntu Hardy 8.04, have you already considered an update path into Pacemaker based clusters? If you have, would you go for Pacemaker/OpenAIS (The New Technology) or stay with Pacemaker/Heartbeat (Will now be supported by Linbit for an undetermined period of time).
Please, let me know.
Thank you 
.
PS. This is just something that I wan’t to know and has nothing to do about with the Ubuntu Cluster Stack Goals!
February 22nd, 2010 at 8:05 PM
Hi,
we have a lot of heartbeat installs running, some of them mostly for master databases with drbd, and at least 6 installs with heartbeat+ldirectord+ipvsadm (some of them with portchannels/bond mode 2)
And yes, we are evaluating the change from standard HB to pacemaker…which needs to be done this year, after lucid is released.
This is all onsite, and no external provider…
regards,
\sh
February 22nd, 2010 at 8:53 PM
We’re migration some old Piranha based load balancers on RHEL3 to Ubuntu Lucid.
We’re going for Pacemaker/OpenAIS/ldirectord/ipvsadm controlled by Puppet so we can easily deploy pairs of load balancers on commodity hardware with minimum fuss.
February 22nd, 2010 at 9:26 PM
@Omahn, Hi there. Do you have a recipe integrate Pacemaker…. into Puppet? It would be awesome to take a look at
February 22nd, 2010 at 9:50 PM
[...] post: RoAkSoAx's Weblog » QUESTION: Cluster Stack! Galley Slaves: A Serious Question About the E.U.Cameron's first response to the bullying [...]
February 24th, 2010 at 8:14 PM
[...] See the article here: Andres Rodriguez: QUESTION: Cluster Stack! [...]
March 1st, 2010 at 1:02 PM
We’re looking at modifying the one available here:
http://github.com/anchor/ha
I also have a very site specific module written specially for ipvs/pacemaker but I suspect the module linked above will be easier to integrate into other sites environments.
April 9th, 2010 at 3:41 PM
we’ve been running on pacemaker only and are currently deploying it on top of openais/corosync only as it has a superior transition time (node status change).
for a basic task like lvs balancing, the old heartbeat (code) should be easy and stable enough thou…
April 9th, 2010 at 3:41 PM
@http://github.com/anchor/ha – you should consider using the crm shell! it makes things much easier