The sysadmins puppet sets up a bunch of handy daemons. I'll need to figure out what of that is applicable for RAID-less systems like our DL120G7's, and get that installed.

Given what's going on on the HP's under load now, we should get this in place. I'm reading up.

[root@relabs08 current]# hpacucli controller all show Error: No controllers detected. So I don't think the ACU stuff will be helpful!

The hp-health stuff is kinda neat - it's a nice command-line interface to the machine-status information available through IMPI, including the IML, a log of maintenance information. It also has hpuid, which allows turning the host's blue light of need on and off. That could be handy. hpsmh appears to be a full Apache install socked away in /opt/hp. I doubt that's useful, and indeed playing with it on a system where it is installed shows it has no information and consumes a lot of userland resources. No thanks! So, I think this isn't useful, but I'll wait for feedback from others in infra.

Related - bug 753804. Ashish suggests installing and monitoring the things that are useful, which is to say, hp-health.

And in bug 779487, it looks like the hosts aren't set up correctly, so maybe hpacucli *will* be useful. I'll investigate.

Nope, hpacucli doesn't recognize these controllers (which, I think, are just basic SATA controllers like you'd find on a retail mobo). I've got a patch ready that installs and starts the HP utilities, and adds check_hpasm and check_hplog checks. It looks like check_hplog would catch (but not fix) the problem in bug 779487: [root@relabs08 ~]# /usr/lib64/nagios/plugins/check_hplog -t l WARNING 0001: POST Error: 1785-Drive Array not Configured

The HP RPMs are partially i386, which is fugly but that's what you get from 3rd party binaries. They also don't correctly require libgcc.i686, so I've added that separately. I consider that a packaging error, so hiding it in the packages::hp_health class is OK, but let me know if you disagree.

Comment on attachment 653898 [details] [diff] [review] bug773648.patch (since catlee's out this week)

Comment on attachment 653898 [details] [diff] [review] bug773648.patch Review of attachment 653898 [details] [diff] [review]: ----------------------------------------------------------------- Assuming that check_hpasm and check_hplog are imported and have proper licenses, I skipped them. ::: modules/packages/manifests/hp_health.pp @@ +1,2 @@ > +class packages::hp_health { > + case $operatingsystem{ A nit. Could you use $::operatingsystem add a space before { here?

I'll fix that whitespace error in all of the packages/maniefsts/**.pp that it occurs in, in a second commit flagged with this bug ID.

http://hg.mozilla.org/build/puppet/rev/8ce0d2390193 http://hg.mozilla.org/build/puppet/rev/7ffaa0895cca - whitespace