cyborgworkshop.org » disk

Ubuntu unstable on ESXi

jason — Fri, 21 Nov 2008 21:18:12 +0000

I’m having some serious stability issues with all of my ubuntu VMs on ESXi. After a period of several days, they have all locked up with a message about the mtp driver having read errors, or something along those lines. The short summary is that it looks like the scsi disk (emulated by ESXi, actually an iscsi LUN) becomes unavailable and lots of sense key errors are generated. The VM goes to 100% CPU and I have to reboot the entire ESXi server to recover. In the meantime, my SAN, Fedora, Solaris and Windows VMs never skip a beat. I’ve rebuilt the VM several times with different version of ubuntu in the 8.x line and different patch revisions, but have finally gave up. I’m back on Fedora and we’ll see if it remains stable.

Quck command to find which filesystems are greater then 90%

jason — Tue, 22 Apr 2008 14:00:31 +0000

While I have a good monitoring system setup, some of my nodes can’t be monitored and require manual checkups. To speed things up, I use a little bash script to log into each host and check a few things out. This snippit below walks through each host in my host file and prints out file systems that are at greater then 90% utilization.

for i in `cat /etc/hosts | grep -v localhost | awk {‘print $3′}`; do
echo $i;
ssh $i “df -k | sed 1d | awk ‘\$5>90 {print}’”;
done

Getting the number of used inodes

jason — Mon, 21 Apr 2008 14:00:57 +0000

I won’t go into the details of what inodes are and why they are important, but when you run out of them, you can have all of the free disk space in the world and not be able to do a darn thing with it. An easy way to see how many inodes are being used on a linux system is with the df command

df -i

This will show you the number of inodes allocated to a file system and the number that are free. Likewise in solaris, the command is df but you use a different option to get the same information.

df -o i