Further to my comment;

http://www.debian-administration.org/polls/64#comment_15

This weekend the PE2650 box had another I/O glitch. Which causes the root disk to become read only (Debians default stance for root disk problems, and very sensible I think).

DELL asked me to install Windows in order to extract the Adaptec Controllers internal log. I'm assuming this means that if we had paid upteem zillion for RHEL we'd still be finding ourselves with second rate diagnostic tools.

Of course the fault could be kernel support for the PERC 3/Di, which might be better in RHEL.

The fault is depressingly similar to;
http://lists.us.dell.com/pipermail/linux-poweredge/2005-May/020931.html

Its about this time I could do with knowing if anyone is successfully running Debian Sarge on a PE2650 with PERC 3/Di, and 2.6 kernel, with write cache enabled. I suspect the exact PowerEdge server model is irrelevant, but the PERC 3/Di probably isn't.

Without a reliable way to reproduce the problem (the box survived extensive stress testing of the RAID subsystem), but just glitches the I/O after about 3 or 4 weeks of up time, it is very hard to figure out if the problem is fixed.

For example I'd gladly downgrade the kernel to 2.4, if only I could be sure that had fixed the problem, without having to put the box back into production and waiting a couple of months for my pager to go off.

Sounds like DELL will replace the RAID controller, cables and disks (urm surely replacing the whole box would be simpler and more reliable), but of course it may be a kernel driver issue. But without a way to reliably reproduce the problem.....

DELL support have been very helpful, but I think I've had enough of Adaptec RAID technology.