Unsolved
This post is more than 5 years old
1 Rookie
•
15 Posts
0
1040
May 3rd, 2017 19:00
Vplex with Vnx in backend performance slow
Hello ,
I need some advise .Customer is using Vplex with VNX as backend storage .
The environment is facing huge performance issue every week that is impacting the production . ISAP application is used and hosted on VMs The SAP engineer ran application tests and statpack report from the server, which points to Log File Sync waits caused by very slow I/O .He also provided the details of the lun that has latency of ~30ms .
I wasn't initially involved in the troubleshooting , and the storage engineer who worked on it has left . As a workaround he changed the Lun policy to highest available tier that brought down the response time . Last week customer again encountered high response time .There was a backup job that was running in parallel with application jobs. When the backup job was cancelled there was significant improvement in the response time. As per SAP team this happens every monday and going on since past one month .
Per the storage engineer update he promoted the luns to extreme performance tier to bring down the response time
Has anyone encountered similar issue ? I am not sure how promoting the luns to extreme performance can stop the occurrence .
Also, let me know the things I can check from my end to know whether the issue is actually from storage .Customer need a permanent fix .
Thank you .
kelleg
4.5K Posts
0
May 10th, 2017 08:00
If you have access to monitoring the VNX's performance (using the Analyzer Archive files - called NAR files) or using the Unisphere GUI and monitoring the Real-Time performance (System/Statistics for Block/Real Time) you may be able to see if there is overloading on the disks on the array. Check the Help for "Performance Monitor" - that provides a lot of information on using the built-in tools on the VNX.
On support.emc.com you can find a number of Best Practice documents for VNX that provides information on the best configuration recommendations.
What you describe sounds like the disks in the Pool that contains the LUN with the performance issue are getting overloaded when the backup runs. You could try to move the backup to a different time.
A last option would be to open a case with EMC specifically for a performance issue on the VNX - this would direct the case to the performance team who can take a look at the array's performance. See:
KB https://support.emc.com/kb/335319 VNX/CLARiiON: How to gather the necessary information for a VNX performance analysis (User Correctable)
KB https://support.emc.com/kb/473729 CLARiiON and VNX: How do I find information about Navisphere/Unisphere Analyzer or performance issues?
glen