Playing around with some HP servers that are going live soon, I loaded
up 2.6.5-mm1 and all filesystems are ext3, the disks are mirrored 36Gb
SCSI using the cciss-driver. After the machine had been exceedingly
wellbehaved I decided to have fun and did this:

formail -s mail < big.mail.box

and thus injected approx 20k mails into postfix. The machine stayed
responsive and well behaved throughout, but the kernel log caught a
few of these:

Badness in elv_remove_request at drivers/block/elevator.c:253
Call Trace:
 [<c022dc31>] elv_remove_request+0x92/0x94
 [<c023af06>] do_cciss_request+0x290/0x2bf
 [<c0235212>] as_remove_queued_request+0x70/0xf9
 [<c011982a>] scheduler_tick+0x17f/0x4b3
 [<c0234721>] as_find_next_arq+0x63/0x76
 [<c02355d1>] as_move_to_dispatch+0xe8/0x1c4
 [<c02a89b1>] schedule+0x445/0x85c
 [<c02357dd>] as_dispatch_request+0x130/0x2b1
 [<c02364aa>] as_work_handler+0x78/0x7a
 [<c012d20c>] worker_thread+0x1a2/0x245
 [<c0236432>] as_work_handler+0x0/0x7a
 [<c0119b5e>] default_wake_function+0x0/0xc
 [<c0119b5e>] default_wake_function+0x0/0xc
 [<c012d06a>] worker_thread+0x0/0x245
 [<c01307c2>] kthread+0x9c/0xa1
 [<c0130726>] kthread+0x0/0xa1
 [<c010225d>] kernel_thread_helper+0x5/0xb
 
The config used is attached.

mvh,
A