Karl Katzke

Mm, alphabet soup.

I hadn’t looked into the cluster world since the last time we configured our cluster for SuSE 10.3 last year. Major groundswells, sinkholes, and shifts seem to have opened up … to the point where things aren’t even called the same anymore.

The big difference is that EVMS is apparently deprecated — even though it’s still listed on the linux-ha projects list, according to the EVMS wikipedia page, development was stopped sometime after 2.5.5 was released in Feb06. That’s a long time ago in internet years.

To get my own brain in the right place: Pacemaker split off from Heartbeat in order to give the whole architecture better … architecture. Pacemaker’s the cluster manager, in the place of the previous clm. OpenAIS/Corosync is one engine (Corosync is the engine itself, OpenAIS is the service plugins to the engine) and Heartbeat is another.

Now… according to the OpenSuSE Mailing List, EVMS is being officially abandoned in place of clvmd, openais, and pacemaker. And indeed, the packages are available in OpenSuSE 11.1, although as I noted a few days ago, the SLES11 High-Availability Extension … isn’t available. If that isn’t irony, I dunno what is. Word on a release date is soon ™… and that was last month. From the current buzz, it went from being “weeks” last month to “it’s in release candidate” this month to “it might be available in at least a month or so”… which means it’s still vapor at this point unless you’re lucky enough to be a Novell Beta partner. We’re not, unfortunately, no matter how good of guinea pigs we may be.

Idle Speculation: It’s probably a bear to provide support on to clueless ex-Netware graybeards, and developing support documentation and scripts for phone monkies takes time and money. My guess is that they’re boiling it down a bit more before they release it to the general public because it just isn’t quite ready for prime time use.

I got the packages all set up and installed on my test mule yesterday… Monday I’ll hook up a second box and see how things go. For now, we’re running everything critical that we can on SLES11 (ex: our new database server, which I just brought into production today)… but running everything else on OpenSuSE. We’ll slowly upgrade from 10.3 to SLES11 where we can, and OpenSuSE 11.1 on the hosts that need to be clvm-aware.