In case you haven’t noticed, we no longer manage clusters with XML

May 20, 2010

While Yves presents a valid approach for managing NDB with Pacemaker and the anything Linux-HA resource agent (a generic wrapper for any daemon based application), the XML configuration shown is likely to have people running away screaming. This is how we had to do things back in the Heartbeat 2 days, which meant that as a cluster admin you would permanently run around with a loaded cocked shotgun with the muzzle pressed firmly against your foot.

Those days are long gone. Please don’t do this anymore. You also no longer need to use the low-level tools such as cibadmin or crm_attribute.

We have the crm shell for those things now. So instead of

crm_attribute -t crm_config -n stonith-enabled -v false
crm_attribute --attr-name symmetric-cluster --attr-value false

we now simply write

crm configure property stonith-enabled=false symmetric-cluster=false

Can’t remember a property name or supported value? Interactive mode and tab completion to the rescue:

# crm configure
crm(live)configure# property sto<tab>
stonith-action=   stonith-enabled=  stonith-timeout=
crm(live)configure# property stonith-enabled=<tab>
stonith-enabled (boolean, [true]):
Failed nodes are STONITH'd
crm(live)configure# property stonith-enabled=false symmetric-cluster=false
crm(live)configure# commit

A bit more convenient, isn’t it? Now let’s get rid of dealing with XML. Instead of a horrific avalanche of tags and attributes that we then have to feed into the system with cibadmin, let’s do this!

# crm configure
crm(live)configure# primitive mgmd ocf:heartbeat:anything \
	params binfile="/usr/local/bin/fake_ndb_mgmd"
crm(live)configure# primitive ndbcluster ocf:heartbeat:anything \
	params binfile="/usr/local/bin/fake_ndb_cluster_start"
crm(live)configure# primitive ndbd ocf:heartbeat:anything \
	params binfile="/usr/local/bin/fake_ndbd"
crm(live)configure# clone ndbdclone ndbd \
	meta clone-max=2
crm(live)configure# location loc-1 mgmd inf: testvirtbox
crm(live)configure# location loc-2 ndbcluster inf: testvirtbox
crm(live)configure# location loc-3 ndbdclone -inf: testvirtbox
crm(live)configure# order ordered-set-1 inf: mgmd ndbdclone ndbcluster
crm(live)configure# commit 

And of course, we get tab completion and online help there too. Give it a shot!

Ubuntu 10.04 with full cluster stack support

May 3, 2010

Just in case you haven’t noticed, Ubuntu 10.04 “Lucid Lynx” has arrived. I upgraded my personal box over the weekend, and am happy to report that things went pretty flawlessly for me. So I’m now a happy lucid user.

What’s more important for the ever-growing cluster community is the fact that this LTS release comes with built-in support for the Pacemaker cluster stack — and indeed, it’s the first commercially-supported distribution that comes with support for both the Heartbeat and the Corosync cluster messaging layer. And it’s easy, too!

Here’s how I install the Pacemaker stack with Heartbeat (which I prefer) on an Ubuntu box:

aptitude install heartbeat pacemaker

And if you choose to go with Corosync, here’s how.

aptitude install corosync pacemaker

… and you’re good to go!

Credit for Ubuntu cluster stack packaging goes to Ante Karamatić (ivoks) and Andres Rodriguez (roaksoax), along with the rest of the ubuntu-ha team. Their work builds on Debian packaging by Martin Loschwitz, Guido Günther, Frederik Schüler and others.

MySQL master/slave support merged into Linux-HA

April 21, 2010

MySQL replication support for the Pacemaker cluster manager (the stuff that we explained in this webinar) has made it into the Linux-HA resource agents default branch. If you are interested in testing — and you should! — please read the extended announcement. Feedback is extremely welcome on the linux-ha-dev mailing list.

We are expecting to release this as part of resource-agents 1.0.4, in late May/early June.

Upcoming webinar: Oracle high availability with DRBD and Pacemaker

April 20, 2010

We’re sticking with databases for our current round of webinars. Up next is an overview of Oracle high availability clustering on Linux.

In this 45-minute presentation, we will show you how to quickly and easily configure an Oracle database with an associated TNS Listener in a failover configuration, how to monitor both your database and your listener for failures, and how to have Pacemaker automatically intervene and recover from outages.

Brought to you in association with our friends over at Novell, we’ll showcase Oracle in combination with SUSE Linux Enterprise High Availability Extension.

This webinar is scheduled for April 28, 2010 at 1500 UTC.

Vodpod videos no longer available.

As all of our webinars, this one requires registration — but we’ve made things easier for you. Just fill in the registration form right here and you’re good to go!

By the way: if you want to be notified of future webinar announcements, you can add the categorized RSS feed for webinars to your feed reader. This is the RSS 2.0 URL:

See you at LinuxTag 2010

April 12, 2010

I just got word that a talk about iSCSI High Availability with Pacemaker and DRBD, which I submitted for LinuxTag 2010, has been accepted by the selection committee.

The presentation is entitled Storage Done Right: Building a Resilient, Distributed, Highly Available Open Source iSCSI SAN and I will talk in the Storage track, in English. Questions in German, of course, won’t be a problem.

As the conference organizers have asked speakers not to publicize the temporary schedule (for obvious reasons — hey it’s temporary), I can’t give the exact time and location yet. But if you want to hear about how you can replace your six-figure SAN with something much more open, much less locked-in and much less expensive, plan a trip to Berlin in early June.