10-25-2011 08:12 AM
We have 3 servers using 1741 CNA's (in a Dell Blade chassis) that connect to a CX4-120 using FIber. Per EMC, we are experiencing performance issues that are related to QFULL occuring at the SP Ports. EMC is stating to lower the throttle execution on the HBA's to 32 to help eleviate the issue.the When we go into the advanced configuration on the CNA's, the queue depth field is grayed out. Does this mean we cannot change the queue depth? My understanding is that this is how we can lower the execution throttle on these HBA's.
Also of note is that the fabric is wide open (i.e. everyone can see everyone). Could this alone be causing the QFull problems at the CLARiiON level? I've been working with the Admins at the location having the problems to fix the zoning, but don't think this would be causing the problems, but could be wrong. The CX4 is barely being hit. It is peaking at just 300 IO's total and force flushing is not occuring at this point. The array is healthy with no problems. The fabric contains 2 arrays, the CX4 (FLARE 30) and a CX300 (FLARE 29) and about 25 servers in total. Anyone help out with this?
10-25-2011 09:00 AM
Qdepth feature and performance will be implemented in the next year releases of driver. You can't change the qdepth as of today. Which OS environment are you currently testing the 1741. Also are you using the 1741 in FCoE?
Below is the URL of the performance analysis for the EMC Clariion which also mentioned about Port Queue Full Count discussion: