Unsolved
This post is more than 5 years old
46 Posts
0
1041
Clariion SP util unbalanced. Where do I go from here
Hi have a Flare 29, CX4-960 with 400GB EFD drives. A test host has 26 luns balanced behind SPA and SPB both in terms of number of LUNS and IO load of those luns.
Yet the load on SPA is consistently 30% higher than SPB. I only have one test Solaris host on the array running PowerPath 5.2 SP1. The host is registered as Clariion open failover mode 4 (ALUA)
The host is zoned via two HBA's to SPA0/1 and SPB0/1. We also see the 1 ports doing 30% less than the 0 ports.
I want to trouble shoot this to make the Util and load equal on both SPs and all ports. Can someone suggest where do I go from here?
nandas
1.5K Posts
0
July 21st, 2010 07:00
Are the IO characteristics and pattern the same to all the 26 LUNs? What type of load/data is it?
Are you using any other layered application like snapview, sancopy etc?
How the Cache is configured on SPA & SPB? Is the amount of cache same on both SP?
Anyway - I 'll suggest to collect Navi Analyzer data and start analyzing the same.
My 2 cents
Sandip
BrentRuffridge
40 Posts
1
July 21st, 2010 07:00
One other suggestion is to try different load balancing policies within Powerpath. Default is CLARopt, but you could try Least Blocks or Least I/O to see if it changes the load on your SP ports.
kelleg
4.5K Posts
1
July 21st, 2010 14:00
Have you tried failover mode 1 instead of 4? With PowerPath and load balancing it should be the same, but try 1 and see what happens.
Also - how is the cache set for the individual EFD LUNS - the standard recommendation is to turn off both read and write cache on EFD LUNs unless you have a very specific application that you know will benefit from turning on cache on the LUN.
What are you using to measure load - host based tools or Analyzer? Are you looking at IOPS or Bandwidth? What is your test suite?
glen
Allen Ward
2.1K Posts
1
July 21st, 2010 14:00
How are your HBAs zoned? That isn't clear from your description...
OR
If it is the latter, I would check the settings on the HBAs to make sure they are the same and I would check the switch ports for the host connections to make sure you aren't getting errors on one of the HBA connections.
NeilHopwood
46 Posts
0
July 22nd, 2010 00:00
Wow! Lots of tips and help here guys! Thanks.
In response to your questions:
The IO pattern for all the LUNS is pretty much the same. There is a ZFS built on top of them.
The PowerPath policy is set to CLROpt but I will change to Least blocks and monitor
HBA 0 is zoned to the 0 ports HBA 1 to the 1 ports and there are no errors reported on the switches for HBA or SPs.
Read and write cache is disabled for all luns as per EMC recommendations but we are about to switch that on to see what happens.
I am starting to look at the host also, perhaps one bus is working harder than the other hence the imbalance but I doubt this would cause the SP imbalance.
Thanks again