Excessive number of system crashes on Sparc server

[guido-pe]

Hi,

Lately, I have been seeing a disturbingly large number of system crashes on my Linux/Sparc based server.

A number of years (5 or 6-ish) ago, I bought myself a used Sun Netra t1 AC200 UltraSparc II 500 Mhz server on Ebay, put it in a rack, put Gentoo Linux on it and started using it as a server. This had worked quite well for a long time, uptimes were measured in years, as expected from a Linux system, and downtimes were (usually) because of kernel upgrades or physical server moves.

About two months ago, the machine crashed hard and wouldn't boot anymore either. Apparently I had made a mistake in the initrd for the new kernel I had prepared some time prior... Anyway, I took this opportunity to boot the machine from CDROM and upgrade it from its ancient 2.6.31 (I think) kernel to a brand new 3.5.7 version and massaged its initrd until I was sure it would boot reliably.

Unfortunately, since then I keep seeing system crashes about once every two weeks. These crashes seem to often coincide with my full backups - I am using duplicity for backups, and I have configured it to do incremental backups twice per day plus full backups once every 15 days. The backup is running over OpenVPN and SSH. The combined load caused by this migt have something to do with the crashes, especially since this time around, it coincided with what from the logs looks like a high-concurrency brute-force attack on my SMTP server...

Does anybody have any idea what could be happening here or how I could debug this further?