March 20th, 2008

Ceci n'est pas une personne.

(no subject)

I hate to bad mouth Intel products, as I love Intel, but...

It looks like we might have found out why our HP Proliant DL360 servers have been crashing. We're using 64bit Xeon quad cores procs. The servers with stepping level 7 are stable. The servers with stepping level 11 are crashing with NMI errors. This is running under RHEL4.

Anyway, if you are running Quad Core Xeons under Red Hat, you might want to look in /proc/cpuinfo and make sure you are not using procs at stepping level 11.

I'm not sure how these different apps interact with the proc differently, but the MySQL servers are not tickling this bug, only the Apache/PHP servers. Maybe the php code is doing some funky floating point stuff.

Anyway, that's your SysAdmin PSA for the day.