Gentoo openMosix stability issues

Bartmanw · n00b Joined: 10 Dec 2003 Posts: 10

At our university (Ghent), we are running an openMosix cluster with Gentoo. Our setup is one server + 33 diskless slave nodes. For some time, we were running the official openMosix 2.4.26 kernel (sys-kernel/openmosix-sources-2.4.26-r1).

Unfortunately, our server motherboard failed, and we decided to buy a brand new server. However, it has an Intel ICH7 SATA controller, which is only supported on kernels>=2.4.29. So, we tried the Gentoo openMosix kernels (2.4.30-r3 and 2.4.32), but both of these are very unstable. As long as openMosix is not running (i.e. /etc/init.d/openmosix is not started), everything runs fine. With openMosix running however, I get frequent kernel oopses and panics. Also, running anything with mosrun gives a segmentation fault in the execve() call.

I'll be very, very grateful to anyone who can help me on this, as we currently have 34 computers just collecting dust

gustafson · Guru Joined: 12 Aug 2004 Posts: 430

Have you considered trying to download a open mosix 2.6 kernel version and installing yourself? Its beta but may be worth a try given your situation.

I have considered it several times but suspect the effort would be a bit premature. If you have success I'd love to hear about it.

Bartmanw · n00b Joined: 10 Dec 2003 Posts: 10

Well, I tried 2.6.15 kernel, but unfortunately it has the same stability issues (i.e. kernel oopses). I wouldn't say it's beta already, much important functionality is still missing.

I almost don't want to say this, but I installed MOSIX (the non-open mosix) 2.4.32, and this works flawlessly. It turns out to be almost perfectly compatible with openMosix. I just needed a little kernel hack to link /proc/hpc to /proc/mosix, and now I use a MOSIX kernel with openMosix userspace tools and openMosix init-script.