On Wed, May 28, 2008 at 07:10:08PM +0300, Pasi Kärkkäinen wrote:
> 
> > > 
> > > Basicly those "Nop-out timedout" errors keep showing up all the time when
> > > there is IO going on.. and if I have "dd if=/dev/mpath of=/dev/null" 
> > > running 
> > 
> > You can expand the timeout to a higher value? 30 seconds ? Also you might
> > want to limit the node.session.queue_depth to a lower value as well.
> > 
> 
> I tried this.. doesn't seem to help much. I still get the same errors. 
> 
> I'll try limiting queue depth too.. 
> 

default queue depth is 32. 

I ran:
echo 8 > /sys/block/sdc/device/queue_depth
echo 8 > /sys/block/sdd/device/queue_depth

and re-ran the "dd test". Same problem. Log entries:

iscsid: Nop-out timedout after 10 seconds on connection 14:0 state (3). 
Dropping session.
iscsid: connection14:0 is operational after recovery (2 attempts)

then again it seems I get these errors less often now.. (with a smaller queue 
depth).
So it seems to help.. 

I'm not totally sure about this, but it could be that sometimes when I can see 
the "io stall" (with iostat) I also get that Nop-out timedout.. and sometimes 
not. 

With a smaller queue depth it just stalls, but with a bigger queue depth it 
also 
drops the session (more often).


Results from the "dd test" with noop_out_timeout of 30 seconds and queue depth 
of 32:

iscsid: Nop-out timedout after 30 seconds on connection 18:0 state (3). 
Dropping session.
iscsid: connection18:0 is operational after recovery (2 attempts)
kernel: sd 20:0:0:0: SCSI error: return code = 0x00020000
kernel: end_request: I/O error, dev sdd, sector 13510024
kernel: device-mapper: multipath: Failing path 8:48.
multipathd: 8:48: mark as failed
multipathd: promise_test1: remaining active paths: 1
iscsid: Nop-out timedout after 30 seconds on connection 18:0 state (3). 
Dropping session.
iscsid: connection18:0 is operational after recovery (2 attempts)
multipathd: sdd: readsector0 checker reports path is up
multipathd: 8:48: reinstated
multipathd: promise_test1: remaining active paths: 2
multipathd: promise_test1: switch to path group #1

So hmm.. it looks like lowering the queue depth helps with the session drops 
while increasing 
the noop_out_timeout doesn't make much difference.. 

Or actually, it could be that increasing the noop_out_timeout makes the
stalls happen less often.. hmm:)

Thanks for the help/comments!

-- Pasi

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"open-iscsi" group.
To post to this group, send email to open-iscsi@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/open-iscsi
-~----------~----~----~----~------~----~------~--~---

Reply via email to