On 14.04.2010 10:15, Darren Mansell wrote:
I've been a long-time user of Ubuntu and with Canonical pushing Server
into the Enterprise I really want to see Ubuntu come up with a good
cluster stack. In fact my next HA project is being based on Lucid with
good results so far.
Note that I do not represent Canonical and I don't know what are their
plans. Cluster stack in Ubuntu is mostly community driven effort and as
such has its advantages and disadvantages (most noticeable is probably
lack of time to polish everything).
* Stay with RHCS??
When I've looked into RHCS previously, it's been a little too far off
the mainstream of Linux-HA so it's difficult to get documentation from
anywhere but Red Hat and it now looks like they are going more in line
with Pacemaker and Corosync by employing Andrew Beekhof.
Don't forget we are talking about Ubuntu 10.04, which is a long term
supported version. RHCS is part of Red Hat Enterprise offerings and it
will be supported by Red Hat for some time. Therefor they will most
probably keep it at least on maintenance level of support
We are aware of current development of unified cluster stack and, even
tough we are small part of that project, we try to participate. For
Ubuntu 10.04 we will provide separate PPA which will have updates for
this new cluster stack:
https://edge.launchpad.net/~ubuntu-ha-maintainers
* Pacemaker: Too unstable.
Why is Pacemaker too unstable? It's used for the HA component of SLES
and I see lots of people using it on lots of different distros with success.
This wasn't remark about software usability, but about development. Also
notice that 'pacemaker' is our term for the new cluster stack (yes, I'm
aware that it still unnamed :). So, this new cluster stack is awesome,
but it also has rapid development (notice number of changes in
heartbeat/cluster-glue/cluster-agents). That's the reason why we will
provide stable pacemaker/corosync in universe and rhcs in main for time
being.
* Heartbeat: Linbit is trying to push users to use Pacemaker +
Heartbeat 3.0 to provide an update patch from Heartbeat 2.x
I don't think Linbit are trying to push anyone on to Heartbeat 3.0, they
are supporting Heartbeat as it was effectively unmaintained and it's
another source of revenue for them to be able to support users with
mission-critical DRBD + Heartbeat style clusters. I think most would
agree that Corosync is the way forward as a cluster messaging layer.
That's correct.
I think Heartbeat 3.0 should be low-priority for Ubuntu-cluster-stack.
Corosync and OpenAIS are where it's all happening now.
And it is. Our goal is pacemaker/corosync in main with optional
heartbeat in universe. But this won't happen in this release. We had
plans to move it into main, but for lots of reasons that didn't happen.
* What about LoadBalancing.
* keepalived for software
* If we use this one, we need to support bonding out of the box
ldirectord is also fine, it works and is an interface to LVS, just as
keepalived is.
ldirectord will be available, but keepalived is recommended solution.
It's in main and it is actively developed while ldirectord depends on
lots of perl modules which aren't in main.
Thank you for sharing your thoughts. If you have time, you could help us
and test lucid pacemaker cluster:
https://wiki.ubuntu.com/ClusterStack/LucidTesting
Best regards
_______________________________________________
Mailing list: https://launchpad.net/~ubuntu-ha
Post to : [email protected]
Unsubscribe : https://launchpad.net/~ubuntu-ha
More help : https://help.launchpad.net/ListHelp
