This morning as I continue to work on my new server, I ran in to a serious problem—the /var/log directory was growing at an alarming rate! Both the messages and warn logfiles were over 4Gb each, after only about five hours of operation. The logs were filling with kernel messages that looked like this:
kernel: FAILED
kernel: status = 1, messages = 00, host = 0, driver = 08
kernel: <6>sd: Current: sense key: Illegal Request
And this was showing up dozens of times per second. Thinking I was either on the fast track to a kernel panic or just filling up my shiny new 750Gb (RAID-1) setup, I quickly googled for an answer and found it here. I tried to leave a “thank you” comment, but his comment code is broken. I’m trying to send a trackback instead.
In any case, the short answer is as follows:
cd /var/lib/dkms/megaraid_sas/v00.00.03.16/source
patch < patches/sles10-ga.patch
If you are asked about using the -R option, answer yes.
dkms build -m megaraid_sas -v v00.00.03.16
dkms install -m megaraid_sas -v v00.00.03.16
Reboot your machine. You will no longer have kernel issues with the RAID controller. You may have a different version than I do (03.16), but it should still work.
Be sure and visit Jonathan Delgado’s blog post about this problem for many more details.
I’m currently building a new server for our colo facility to host streaming video. It’s a nice machine, quad-core Xeon processor, 4Gb of RAM, 750Gb of RAID-1. When compiling PHP I kept running in to a problem with the gd library configuration. The shared object files are installed into /usr/lib64 (because it’s a 64-bit operating system), but the header files are installed into /usr/include and are not being found.
If you tell the configure script to --enable-gd=/usr it balks about not finding the libgd.(a|so) file. If you rerun configure and use --enable-gd=/usr/lib64 it finds libgd.so, but then doesn’t pick up the header files.
Turns out the problem is pretty simple to solve: add --with-libdir=lib64 to the configure script and it will fix everything. It’s just not obvious when reading the configure help, especially when you come from a primarily 32-bit background.
Yesterday I took down my machine for two reasons. One, I was running low on hard drive space (I store a lot of CD and DVD ISO images for the office). Two, my numpad stopped working. I have no idea why. I plugged in a different USB keyboard and its numpad didn’t work either.
Anyway, I took the machine down, added the drive into the third SATA hookup (I have two drives in there already, with a maximum of four), and started everything back up again.
I couldn’t log in after it started up. I almost panicked, but being the calm, cool, sysadmin that I am, I tailed some logfiles which led me to discover that despite the new drive being in the third SATA port, it got named /dev/sdb, and the drive that was /dev/sdb was now /dev/sdc. That also happened to be where my /home partition was mounted. So I fixed /etc/fstab, mounted /home, and voila!, I could log in again.
After a quick XFS format of the new /dev/sdb1 partition, I was in business. See how easy Linux is? And people say it’s not for everybody…
OpenSUSE 10.3 was released yesterday. Not many people sharing out the torrent yet, though.