We've had a spate of these lately - about 3 or 4 boxes. I have a
feeling we just got a bad run of motherboards, but thought I'd try to
get some feedback.
These are servers running with the Tyan Thunder K8SD (dual core AMD's,
S2882 model). They have three onboard NIC's, two Broadcom gig NIC's
BCM5704C's and an Intel 82551QM which is a 10/100 board. We use the
Broadcom NIC's in this case with the tg3 driver. The kernel panics
happen somewhat at random, though most frequently seen when scp'ing a
large number of files (after a couple gigs they'll crash pretty
reliably). Last night we had a box go down under very little load.
I couldn't copy and paste from the console or scroll up - the one thing
I got that seems relevant is the following:
RIP <ffffffffa0092774> {:tg3:tg3_poll+177{ RSP <ffffffff8045a878>
The tg3 bit is why I suspect something with the network ports. Has
anyone run into this? What's the best way to track this down - is there
a setting where I can save a core file, or do I just need to hook up a
terminal and leave it on 24-7? Thanks for any help.
--
Eric Solomon
Senior Unix Engineer
Clickability
____________________________________
Simplifying Content Management
130 Battery Street, Suite 300 San Francisco CA 94111
Tel 415-575-5125 eric at clickability.com
Fax 415-538-0839 http://www.clickability.com