2007-08-01 15:26:00
Last night's planned change was supposed to last about two hours: get in, install some patches, switch some cluster resources around the nodes, install some more patches and get out. The fact that the installation involved a HP-UX system didn't get me down, even though we only work with Sun Solaris and Tru64. The fact that it involved a ServiceGuard cluster did make me a little apprehensive, but I felt confident that the procedures $CLIENT had supplied me would suffice.
Everything went great, until the 80% mark... Failing the applications back over to their original node failed for some reason and the cluster went into a wonky state. The cluster software told me everything was down, even though some of the software was obviously still running. The cluster wouldn't budge, not up, nor down. And that's when I found out that I rather dislike HP ServiceGuard, all because of one stupid flaw.
You see, all the other cluster software I know provides me with a proper listing of all the defined resources and their current state. Sun Cluster, Veritas Cluster Service and Tru Cluster? All of them are able to give me neat list of what's running where and why something went wrong. Well, not HP Damn ServiceGuard. Feh!
We ended up stopping the database manually and resetting all kinds of flags on the cluster software. Finally, after six hours (instead of the original two), I got off from work around 23:00. Yes... /me heartily dislikes HP ServiceGuard.
kilala.nl tags: work, unix, meh, sysadmin,
View or add comments (curr. 1)
All content, with exception of "borrowed" blogpost images, or unless otherwise indicated, is copyright of Tess Sluijter. The character Kilala the cat-demon is copyright of Rumiko Takahashi and used here without permission.
2007-08-23 00:23:00
Posted by Thomas
Test commentaar