10400 wrong data port utilisation

[lizardius]

hello all!
I have different 3par:F200, 10400, 10800.
All is good with F200.
I have a little strange situation with 10400 storage. This system has 4 nodes, each node connected with one port to first SAN switch, and second port connected to another SAN switch. I have 8 ports connected to servers but only one port is fully loaded.
https://www.dropbox.com/s/u6mowbevpwhpuui/dataport.PNG

Maybe someone know how to fix this situation?

thank you.

[lizardius]

miracle...
one virtual machine overloaded port
issue self resolved after transferring this VM to another node.

[Richard Siemers]

I had similar issues... your ESX admins need to make sure round robin is set on every datastore on every ESX host. If you have 10 ESX hosts, and 10 data stores, they will have to clickity click 100 times or learn to script it. There are commands inside the implementation guide to make this the default for all NEW datastores added after the default is changed.

http://kb.vmware.com/selfservice/micros ... Id=1017760

I suspect what is happening is that you have 1 or more ESX hosts still using the default "Most Recently Used" policy.

[BryanW]

+100,

I had similar issues... your ESX admins need to make sure round robin is set on every datastore on every ESX host. If you have 10 ESX hosts, and 10 data stores, they will have to clickity click 100 times or learn to script it. There are commands inside the implementation guide to make this the default for all NEW datastores added after the default is changed.

http://kb.vmware.com/selfservice/micros ... Id=1017760

I suspect what is happening is that you have 1 or more ESX hosts still using the default "Most Recently Used" policy.

+100, Richard hit the nail square on the head. I have experienced the exact same issue same with the same fix. Our ESX admins now script this for all of our storage types into their ESX autobuilds.
No idea why VMWare would keep most recently used as the default these days. it is akin to hard coding fiber ports to 2GB.

[afidel]

I had similar issues... your ESX admins need to make sure round robin is set on every datastore on every ESX host. If you have 10 ESX hosts, and 10 data stores, they will have to clickity click 100 times or learn to script it. There are commands inside the implementation guide to make this the default for all NEW datastores added after the default is changed.

http://kb.vmware.com/selfservice/micros ... Id=1017760

I suspect what is happening is that you have 1 or more ESX hosts still using the default "Most Recently Used" policy.

You shouldn't have to click, you should just need to change the policy and reboot the host, the PSAP policy change should take effect the next time the storage stack goes through its claim rules unless you've manually changed something (ie if you manually assign a policy it will keep the manual change but if it was set by existing policy automagically it should change the next time things are evaluated, this is my experience with 4.0, 4.1, and 5.0). Also like almost everything else in VMWare it can also be scripted.

[lizardius]

one additional update: one of our virtual disks was presented twice to hosts. first time this volume was presented to few hosts and second time - to host set, which includes old hosts and some new hosts. IMC don't check this situation this situation leads that only one host port was utilized.