On 08/14/2009 07:21 AM, Ulrich Windl wrote:
> On 13 Aug 2009 at 16:34, Hannes Reinecke wrote:
>> Interesting though, both NetApp and MSA level out at around
>> 64 commands. Maybe we should consider changing the default ...
> Our old HP AutoRAID had a recommended queue depth of 8, but I don't know what 
> the
> actual queue size was. At least at that time the IBM deskstar SCSI disk with 
> Linux
> and an Adaptec 2940 on my home PC was using a queue depth of 250 or so ;-)
> Did you have a look at what the SCSI guys do about the queue depth? AFAIR 
> formerly
> you had to specify the depth wehen loading the module/kernel, but then they 
> did an
> dynamic adjustment. Meybe it's like starting with a huge queue, and once the
> target complains, set a capping limit on the queue size. I didn't see source
> recently...

The 64 commands Hannes is talking about is referring to a target level 
queue depth. I think you are talking to a device level one. For the 
target level one, the target tells us how many commands it can take and 
we do not send more than that. For the device level one, we do not know 
the value, so if we send too many the target will normally send a queue 
full error. We do not handle queue fulls right now. We are working on 
making the code the FC driver use to track queue fulls more common (I 
sent a patch the other day where we will lower the queue depth when 
getting queue fulls) so iscsi can hook in the ramp down and ramp up code.

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 
For more options, visit this group at http://groups.google.com/group/open-iscsi

Reply via email to