both paths to equalogic shows same iscsi ip

can you post the output of:

- esxcfg-vswitch -l

- esxcfg-vmknic -l

- esxcli swiscsi nic list -d vmhbaXX (where XX is the iscsi initiator's adapter number (can get it from the storage adapters list in the config tab of the vSphere client).

I also have the same issue with a PS4000 and a PS6000. Both are configured for multipathing but if you look carefully at the activated paths towards a volume, you will soon realize that 2 paths from a host can possibly use the same ip/nic on the storage.

This is completely random behavior and affects random esxi hosts. If you change the storage nic IPs and you refresh the adapters on all you esxi hosts, it just happens that sometimes you correct hosts that had the problem and ruin others.

On PS4000 as a workaround using DELL MEM I used the following command to up the volumesessions (Paths) to a given volume.

setup.pl --setparam --name=volumesessions --value=3 --server=ServerIP

Having 3 paths now to the storage I statistically get at least 2 Paths to a different IP/nic of the storage.  Once I find 2 good ones I disable the third which is redundant.

Benchmarks with IOmeter also follow this practice with better results using Access Specs of 32KB Seq Read. I get about 200+MB/s

Obviously this workaround will not work for PS6000 since max volumesessions is 4 and from the 4 paths I always get 2 with the the same IP/nic of the storage.

I am not sure this is correct but after trying every white paper, removing vswitches, re installing-reconfiguring them in cli (in accordance to TR1049 v1.2) and doing the same with DELL MEM all over again for the last 2 days, i feel that I am missing some info or something is not right with MPIO.

Paths are not picked up and checked for duplication but in a random fashion. Sometimes this results in having double paths to a single nic of the storage. I hope I am missing something here

on a PS6000 when you have a lot of volumes, many Paths and many ESXi hosts accessing data, this effect is not very visible since you rarely need all that BW.by having a couple of duplicated paths might mean that you loose a 1/4 of the unused BW which is not much.

On PS4000 though access to a single volume from a single host where there are 2 paths ending in the same storage nic/IP results in low performance. Actually exactly half of the BW available i.e. 1Gbit. Those speculations are verified by IOmeter.

I was expecting that i missed some info or a configuration step.

I am not sure if we are clear with what me and the fellow user above are describing and if its clear for you. Maybe the picture bellow makes it clearer. Its not a very clean workaround. Its also not full proof if you consider that the Path target ip/nic might change as you said.

I've seen the same issue on my SAN since purchased.  I also have rebuild VMware / recreated vswitches / followed the Dell documentation (Dell EqualLogic MEM Installation and User Guide) and I'm still seeing the same thing as you are.  I have only recently needed this "full" bandwidth for one of my VMs and would like to get this resovled.  Did you have any luck finding a resolution or are you still using your workaround?

The latest firmware with the same setup and new vmware MEM seems to have way better performance on PS4000 PS5000 and my PS6000. I only recently tried performance on a few EQLs and the issues seem to be resolved for me. Try it out.

I am currently on firmware 5.1.2.

is there a big difference if i go to 6.0?

I can confirm this plus you can benefit from the newly provided synchronous replication.

I am at this point running 6.0.1 with MEM for vmware 1.1. After the upgrade to vmware 5.1 I will need to upgrade mem to 1.1.1 (since only that supports vmware esx 5.1) hoping that all stays as is :)

what version number to download and upgrade to. I am on 5.1.2.

6.0.1-L1 or 6.0.1?

what is the recommended proceure to do this upgrade?

do I need to upgrade to 5.2.2 or 5.2.4 before 6.0?

ok I as able to sucessfully update from 5.1.2 to 6.0.1L1 via http. now do I need to update the eqlogic mem module on the esxi servers?