Pascal Haakmat wrote:
[snip]
> In any case the problem does not seem to be with XFS; also, Andre Hedrick
> has been quite vocal on lkml about problems with the block layer/IDE/ATA
> (I think) support in the 2.4 kernel.
>
> I will see if the machine survives the 160 cp processes test tonight. If
> it does, then I would consider the problem solved. Meanwhile thanks for all
> your help.
The machine passed the test with flying colors. It went through 987
iterations of Adrian's script with 160 cp processes for a period of 24 hours
without a single error. These results, again, after applying Andre Hedrick's
IDE patch: without this patch I get "lost interrupt" messages, severe
filesystem corruption, and eventually the entire machine goes down.
So I've restored from backups. Next thing: to vote for the inclusion of
Andre's patches into 2.4 on lkml...
PS: In my last mail I referred to "Andre's script" when I meant "Adrian's
script". Sorry about that.
|