Comment 49 for bug 453444

Hi all,

This bug has been a real pain in my backside. I think I have a reasonable, short-term workaround. I'm *not* suggesting this is a permanent fix for anyone, nor should this be applied to any source trees anywhere.

This is strictly for someone who, like me, is suffering because of this, and wants a slightly better solution than simply killing off logging systemwide till a proper patch is available.

Here's my hack:
--- /etc/init/rsyslog-kmsg.conf.orig 2009-11-17 01:40:07.000000000 -0500
+++ /etc/init/rsyslog-kmsg.conf 2009-11-17 01:42:36.000000000 -0500
@@ -18,7 +18,7 @@
     chown syslog:syslog /var/run/rsyslog/kmsg
 end script

-exec dd bs=1 if=/proc/kmsg of=/var/run/rsyslog/kmsg
+exec dd bs=1 if=`grep -v "CPU[0-7]: Temperature" /proc/kmsg` of=/var/run/rsyslog/kmsg

 post-stop script
     rm /var/run/rsyslog/kmsg

Or, for people who don't like dealing w/ patch files, go to /etc/init, make a copy of rsyslog-kmsg.conf. Then, edit rsyslog-kmsg.conf, find the line near the end of the file that looks like:
dd bs=1 if=/proc/kmsg of=/var/run/rsyslog/kmsg

and replace it with:
exec dd bs=1 if=`grep -v "CPU[0-7]: Temperature" /proc/kmsg` of=/var/run/rsyslog/kmsg

So, all this does is filter out *all* warnings referring to CPU temperature fluctuations (for boxes up to 8 CPUs). Don't look at me if you don't see the warnings and end up flaming out your CPU. :-)

Hope this helps someone until such time that a proper fix is available.

-Mark