MSA Storage
1753543 Members
5082 Online
108795 Solutions
New Discussion

Need Advice on Configuring Dell Blades with QME2572 HBAs

 
DataHammer
Occasional Contributor

Need Advice on Configuring Dell Blades with QME2572 HBAs

SQL 2008r2 EE

Windows 2008r2 EE

5-node cluster (physical)

22 clustered SQL Instances

EVA p6300, 64 disks in 1 group

Dell M1000E Blade Chassis with

Dell 710HD blades

and qLogic QME2572 Mezzanine

 

We are getting stuck IO messages from SQL Server (IO taking longer than 15 SECONDS) when hosts and EVA are under what should be a light IO loads;  total host MB avg at 80K, max at 309K, and IOs avg 1647 max at 5000.  We've done a bunch of testing, ruling out EVA and bandwidth issues at every point in the IO chain. testing has shown that each time we encounter the SQL error there are very high disk queues on the host (1300+). HP cant help us bc we have dell HBAs. Great. HP dosent seem to have a HBA agnostic formula for determining HBA queue depths... Can anyone provide advice as to how to go about determining the appropriate "Execution Throttle" and "LUNs per target" values for our environment?

2 REPLIES 2
giladzzz
Honored Contributor

Re: Need Advice on Configuring Dell Blades with QME2572 HBAs

Hi

You might try and implement this document

http://h20000.www2.hp.com/bizsupport/TechSupport/Document.jsp?lang=en&cc=us&objectID=c02518189&jumpid=reg_R1002_USEN

because it is about connecting windows 2008 to EVA

Regards

 

Press the Kudos star if you liked

 

DataHammer
Occasional Contributor

Re: Need Advice on Configuring Dell Blades with QME2572 HBAs

Thanks for the reply!

 

The problem ended up being that an update package from Dell updated the HBA driver and set the config parameter "Execution Throttle" back to the default of 16 (nice). We did have the doc you linked, but we had pretty much the opposite problem: Host were too heavily throttle, SAN could handle much more traffic. So we needed to increase the value. We set it up to 256, which was effective in openning up the hosts to push more IO. Disk queues decreased dramatically, and throughput went up. 

 

We'd still really like to get some guidance on tuning the HBA parameters for the p6300 specifically. So far we have this

 

http://networkadminkb.com/KB/a180/performance-tuning-the-execution-throttle-of-a-qlogic-hba.aspx

 

The formula suggested here is:

(Maximum Storage Port Command Queue) / (Host Ports) = Execution Throttle

 

But HP support has been unable to tell me what the the Max Storage Port Command Queue is for the p6300. They dont have a formula for setting queue depth either... not too helpfull