BumpInTheWire.com

Sometimes finding something to write about is difficult.  Other times ideas just fall into your lap.

If you use virtualization, do yourself a tremendous favor and run memtest86 or memtest86+ on your server before it goes into production.  We’ve been fighting an ESX server Purple Screen of Death problem for three or four weeks to long now.  Luckily this was a lab server so it was more of an inconvenience than anything.  After looking at a zdump and extracting the kernel log file (vmkdump -l dumpfilename) I was able to determine that it was a page fault error based on the Exception Type 14 that was in the log just before the server puked on itself.  The second dump file I looked at had completely different information so I wasn’t completely convinced it was a memory problem.  With this I opened a support case with VMware support so somebody much brighter than myself could look at the logs and give us a definitive answer.

This is where memtest86 comes in.  The longer the day went on the more I convinced myself this was a memory problem.  So tonight before I left work we threw a memtest86 disk in and bounced the server to let it run overnight.  Sitting here watching the Kansas/Missouri game (which was a real snoozefest until the last 1:50) I thought I’d see if any errors had been detected.

Next time you get that spanking new server all put together and are giddy with anticipation from the thought of installing an OS on it, pop in a memtest86 cd and let it run overnight.  You’ll thank me for it later.