Skip to main content
Start a Conversation

Unsolved

This post is more than 5 years old

Ram_Emc

17 Posts

1041

September 15th, 2014 00:00

LUN service time is very high in CX4, eventhough the throughput is very low

Hi

Apps team reported  an incident on 1 Sep that the batch cycle would suppose to take 1 hr , had taken almost 2hrs to complete.

It is oracle DB with Linux OS using CX4-240 systems with mirror view synchronous across sites.. Oracle Vendor analysed the AWR report and reverted that the log wait event is very high on redo buffer disk which is very slow response from redo disk.

Redo LUN is 20GB from RAID10 pool, when i check the analyser the IOPS of the lun approx below 100, where as the service time is very high almost 250-300ms during 3-5PM.

Where as other luns with heavy load during that period (almost 4000IOPS) the service time is below 30ms.

How do i interpret this info ? EMC support dont find any issues identfied on the storage frame based on the logs.

What are all the possible causes that resulted in high service time?

undertaker98

75 Posts

September 15th, 2014 03:00

If redo luns shared same disks with high load luns, high load luns may occupied a big part of resources.

In this situation you must move redo luns to other disks or you can use QoS Manager to limit high load luns iops or bandwith.

In other words, high load luns generate a big part of IOs and redo luns a small part, all of this IOs queurying to Disks, without priority all IOs are same. In IO queue response time for redo luns are bigger, because redo luns IO rarer met in queue.

P.S. You can create SR in EMC Support with performance problem, not only with software or hardware issues

vigneshkadirvel

13 Posts

September 15th, 2014 04:00

Before going for lun migration, follow below steps

  1. Check in fabric is there any discards, if you have found clear those.
  2. Use  powermt output to check all paths are availble to host for servicing i/o's.
  3. Perform lun migration as undertaker mentioned.

kelleg

4.5K Posts

September 18th, 2014 14:00

Please check KB article KB  91353 on support.emc.com - this explains about an issue where the Response Times are high when the IOPS are low. See the links in the KB for other issues that could also be what your issue is.

For all issues about performance see KB 12289 - this KB lists all the KB's about performance.

glen

Was this post helpful?

View All

No Events found!

Top