Ulrich Windl wrote:
> Hmm, I wonder: If every packet exchange for an iSCSI target re-triggers a 
> timer 
> that prevents NOPs from being sent/expected, the problem under heavy load 
> should 
> go away as well, right? I see little sense to send extra NOP queries when 
> there is 
> heavy traffic for a target. (MHO)
> 
Ach, if it were so easy.
Yes, of course you are correct. And yes, we shouldn't be sending NOPs when under
high load. In fact I hooked into the tcp_data_ready() callback to make sure to
update the timer whenever a new skb is available from the network stack.
And even in any unterminated while() loop in the xmit and recv codepath.

The problem is that I'm _still_ seeing the occasional nopouts. I did some 
analysis
and printed the MIB statistics (ie octects send / received) when the NOP was
being active. And it really _looks_ as if there is no traffic on this connection
during that time. And what's more worrying, apparently we're sending a NOP and
won't get a reply.

I'm still fighting to get a wireshark trace; but dumping two _heavily_ loaded
connections gives you massive traces. So far the disk space has run out before
I got anything interesting noted in them :-(

More updates to follow.

Cheers,

Hannes
-- 
Dr. Hannes Reinecke                   zSeries & Storage
h...@suse.de                          +49 911 74053 688
SUSE LINUX Products GmbH, Maxfeldstr. 5, 90409 Nürnberg
GF: Markus Rex, HRB 16746 (AG Nürnberg)

--~--~---------~--~----~------------~-------~--~----~
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 
open-iscsi+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/open-iscsi
-~----------~----~----~----~------~----~------~--~---

Reply via email to