Kernel stability question

Kurse · Posted: Wed May 03, 2006 2:39 pm Post subject: Kernel stability question

This may or may not be specifically related to Gentoo kernels, as I have not tested with a vanilla kernel, but it seems to me that the last reliably stable kernels were the 2.6.14 kernels. For me, all of the 2.6.15 and 2.6.16 kernels have been very erratic and prone to lockups and crashes.

My hardware has not changed in my servers, so my kernel configs have not changed very much between these versions. They all have standard hardware, nothing really out of the ordinary about them that I would think could cause problems. Never had any problems with any of the kernels prior to the 2.6.14 either. Gentoo has been running on a couple of these for well over 2 years.

Has anyone else noticed or experienced any instabilities like these? Just curious. The couple of Gentoo boxes I have are not mission critical, they are more utility machines, but before I go spend the time and resources to investicate, I thought I would ask the community to see if anyone else may have experiences that are similar.

sundialsvc4 · Guru Joined: 10 Nov 2005 Posts: 436

No. I would suggest that you first look at the hardware. Read the system logs closely. Do a smartctl scan on all your hard-drives. (You can do this while the system is running!)

Oddly enough, when a system is "unstable" it is often not the kernel that's to blame: it's another subsystem, like "X."

Another possibility to consider is that perhaps not all of your kernel-modules are up to date with the rest of the system. You can do something like this ...

After copying your .config file to a safe place as a routine backup of that file...
mv .config foobar
make distclean
mv foobar .config
make && make modules_install && mount /boot && make install && umount /boot

An off-the-wall but serious possibility also is electrical power! A quality UPS-box (uninterruptible power-supply) will provide rock-steady voltage. Many of the devices in a computer run on direct current, and even the slightest drop in power-levels can cause serious and non-reproducible problems. So... does the system crash when someone makes coffee or turns on the photocopier?

bluedevils · Posted: Wed May 03, 2006 4:13 pm Post subject:

I too have had some things break since upgrading the kernel (2.6.12-r10 to 2.6.16-r3) to the point where I had to upgrade more than a few things from being masked. I'm still working things out with baselayout.

Kurse · Posted: Wed May 03, 2006 7:35 pm Post subject:

I dont think I have a hardware issue. If I boot from a 2.6.14 or older kernel, I dont have any problems. Its only when I boot from a 2.6.15 or higher revision that problems occur. There are also several other servers running along side of it, one is Fedora Core 5 (with 2.6.16-1.2096_FC5 kernel), and the other is FreeBSD 6 (Prerelease), neither of them have problems.

The Gentoo server is not running X. It is a standalone utility box, running apache, mysql, and courier-imap with squirrelmail. It is a Pentium 4 3Ghz system with 2GB RAM, adaptec 2100S raid card (RAID0, striped across 3 seagate drives), a sony tape drive, and an nvidia geforce2 card. It is a very standard setup, nothing fancy at all.

I guess I will dig into creating a new kernel config from scratch, I guess I could have missed something along the way.