CX4 Powerpath Lun Issue

Question

Hi, We have a Windows 2003 host connected to CX4 with Powerpath and SQL on it. It was running fine up until middle of Jan, the SQL guys created another DB on this Physical Server and we now get SQL  Lun performance issues and dead path issues see below. We know the CX4 is busy  and  we had some issues with the Database and noticed the some of the DB luns were on SATA so have moved these  to FC. We are now finding the SQL backups, which are maintenance scripts that export the SQL database have now gone from about 25 mins to about 4 hours, and last night we saw lots of event log errors about dead paths and luns We saw the following SQL errors The log for database tempdb is not available Error 9001 severity 21 state 1 Write error during log flush And windows errors where we lost access to P drive for a split second Path Bus 3 TGT 0 lun 3 to xxxxxxxxxxxxxx is dead. So i presume the tempdb on the P drive is timing out and SQL throws a bit of a wobbly as the tempdb dissapears for a split second. My question is should windows and sql be able to handle this latency, or do we need to look at queue depth or some tuning on the HBA or powerpath to take account of this high I/O or is there something that could be done in SQL, Thanks Andy

kingcap100 · Accepted Answer

Michel, Thanks for this, we are going through a process of moving some Luns around, from Sata to FC, and looking to balance the SP's. We will look at the HBA stats as well.

dynamox · Answer

do you have Navisphere Analyzer, what response times are you seeing for the LUNs presented to this server ? If you don't have Navi Analyzer, have you looked in Perfmon at these counters:

PhysicalDisk\Avg. Disk Sec/Write

PhysicalDisk\Avg. Disk Sec/Read

Mich · Answer

The symptoms your are presenting are typical of a performance issue. I recommend you open a case with EMC Support, we can check the basics for errors on the HBA statistics etc and see if there is anything else from the host causing this.

You will probably see Event ID 129s reported from the HBAs as well for disk timeouts. We can narrow down which LUNs they may be as well.

An EMCReports would be required from the server.

Regards, Michel.

EMCReports instructions:

Please download the appropriate version of EMCReports from the following site: ftp://ftp.emc.com/pub/emcgrab/Windows

Execute this tool on the Windows server using an administrator account. For Windows 2008 servers, please right click the EMCRPTS executable and select "Run as administrator".

When you run this tool, you will see a command window start up. The tool runs in that Window. When the tool is done you will see an explorer window open up. In the explorer window is the .zip file that you need to attach to the case in the EMC support site https://support.emc.com using your case number.

The readme file at the FTP location above has more information about the tool if needed.

kingcap100 · Answer

Thanks we will collect this info and we do have Navi Analyzer

VNX

Was this post helpful?