Many administrators just run "dmesg" to check a system for problems and do not bother with its options. But actually it is worth to know about the filtering and output options of the most recent versions (Note: older distros e.g. CentOS5 might not yet ship these options!). You always might want to use "-T" to show human readable timestamps:
$ dmesg -T
[...]
[Wed Oct 10 20:31:22 2012] Buffer I/O error on device sr0, logical block 0
[Wed Oct 10 20:31:22 2012] sr 1:0:0:0: [sr0] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
[...]
Additionally the severity and source of the messages is interesting (option -x):
$ dmesg -xT
[...]
kern :err : [Wed Oct 10 20:31:21 2012] Buffer I/O error on device sr0, logical block 0
kern :info : [Wed Oct 10 20:31:21 2012] sr 1:0:0:0: [sr0] Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
[...]
Now we see that only one of those example message lines was an actual error. But we can even filter for errors or worse ignoring all the boot messages (option -l):
$ dmesg -T -l err,crit,alert,emerg
[...]
[Wed Oct 10 20:31:21 2012] Buffer I/O error on device sr0, logical block 0
[...]
In the same way it is possible to filter the facility (the first column in the -x output). For example this could return:
$ dmesg -T -f daemon
[...]
[Wed Oct 10 19:57:50 2012] udevd[106]: starting version 175
[Wed Oct 10 19:58:08 2012] udevd[383]: starting version 175
[...]
In any case it might be worth remembering:
- -xT for a quick overview with readable timestamps
- -T -l err,crit,alert,emerg to just check for errors
I recently created a simple dmesg Nagios plugin to monitor for important messages with Nagios. You can find it
here.