out of memory in ata_raid_init_request

[Ashley Moran]

I've got a server with an Asus A8V Deluxe board and two Seagate Barracuda 
7200.9 disks plugged into the VIA controller in RAID 1.  Every now and again 
I get errors like this:

Mar  9 04:10:51 don FAILURE - out of memory in ata_raid_init_request
Mar  9 04:10:51 don FAILURE - out of memory in ata_raid_init_request
Mar  9 04:10:51 don FAILURE - out of memory in ata_raid_init_request
Mar  9 04:10:51 don Fg_vfs_done():ar0s1h[WRITE(offset=87667261440, 
length=16384)]error = 5
Mar  9 04:10:51 don g_vfs_done():ar0s1h[WRITE(offset=87667343360, 
length=32768)]error = 5
Mar  9 04:10:51 don g_vfs_done():ar0s1h[WRITE(offset=87667474432, 
length=16384)]error = 5
Mar  9 04:10:51 don AILURE - out of memory in ata_raid_init_request
Mar  9 04:10:51 don g_vfs_done():ar0s1h[WRITE(offset=87668228096, 
length=16384)]error = 5


I posted this to questions@ last time it happens but nobody could offer any 
advice.  Is it hardware-specific, is it significant, and is it another reason 
to switch to gmirror?

It is likely caused by postgres as that would have been doing a data import 
task around the time of the error.  I haven't yet configured postgres to log 
anything useful, but the last entry in /var/log/messages was
Mar  9 04:15:21 don postgres[70314]: [1-1] ERROR:  tuple concurrently updated


Ashley