Re: 2.6.12-rc6-mm1 & 2K lun testing
From: Nick Piggin
Date: Wed Jun 15 2005 - 20:51:37 EST
Badari Pulavarty wrote:
On Wed, 2005-06-15 at 12:02, Nick Piggin wrote:
Yeah, take off GFP_HIGH and set GFP_NOWARN (always). I would be
interested to see how that goes.
Obviously it won't eliminate your failures there (it will probably
produce more of them), however it might help the scsi command
allocation from overwhelming the system.
Hmm.. seems to help little. IO rate is not great (compared to 90MB/sec
with "raw") - but machine is making progress. But again, its pretty
unresponsive.
Anything measurable that we can use to maybe get the chage
picked up and tested in -mm for a while?
Thanks,
Badari
procs -----------memory---------- ---swap-- -----io---- --system--
----cpu----
r b swpd free buff cache si so bi bo in cs us sy
id wa
131 254 34896 31328 2540 4982740 0 0 29 101877 1086 11220
0 100 0 0
149 268 34896 32824 2536 4983712 13 0 42 39505 439 4454 0
100 0 0
135 254 34896 31112 2536 4984768 11 0 20 36233 373 4078 0
100 0 0
130 242 34896 32600 2536 4987364 6 0 161 33626 377 3957 0
100 0 0
153 263 34896 32592 2532 4993560 0 0 14 37124 385 4468 0
100 0 0
144 236 34896 32668 2548 5013148 6 0 154 220366 2360 27530
0 100 0 0
Though it can be difficult to judge performance based on vmstat
when you get these large spikes. vmstat is measuring requests
into the elevator so you see batching and throttling effects. I
would expect requests completing to be more even... your entire
vmstat listing looks like it is averaging about 60-70MB/s - does
this agree with your measurements?
Finally, do you see anything interesting on the profiles?
--
SUSE Labs, Novell Inc.
Send instant messages to your online friends http://au.messenger.yahoo.com
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/