RobBuxton
Enthusiast
Enthusiast

Best Practise for Manage Path Policies

Jump to solution

Hi There,

I'm getting conflicting advice on the Best Practises for Managed Paths.

We are on ESXi version 4.0 connecting to an HP EVA8000. The HP Best Practise Guide recommends setting the Manage Paths Policy to Round Robin.

This seems to give two Active Paths to the optimised controller. See: http://h20195.www2.hp.com/v2/GetPDF.aspx/4AA1-2185ENW.pdf

We employed some consultants and they are saying that the Vmware Best Practise for this solution is to use the MRU Policy that results in a single path to the optimised controller.

So, any idea which best practise is best practise? Does it make a difference?

TIA

Rob.

Tags (3)
0 Kudos
1 Solution

Accepted Solutions
jpdicicco
Hot Shot
Hot Shot

Always go with the storage vendor's recommendation. The VMware recommendation is based on generic array characteristics (controller failover methods, ALUA-capable, etc). The storage vendor's recommendation is based on their performance and compatibility testing. You may want to consider their recommendations carefully, however, to be sure that each point is what you want.

With the 8000, I have run with Round-Robin. It is the most robust pathing option available to you from a performance and failover standpoint, and can provide the most even performance across the ports on the storage controller.

While I have not done specific testing/validation, the last time I looked at the docs, the HP configuration recommended that you setup each I/O to switch ports in the configuration. This adds load to the ESX host, performing the switch across ports, but HP claims their testing showed that this was the optimal configuration. This was the only setting that I questioned in their recommendation.

If you haven't already, make sure you download the HP doc on configuring ESX and EVA arrays. It has several settings that you should configure in addition to the path policy, as well as some scripts to help with making the changes.



Happy virtualizing!

JP

Please consider awarding points to helpful or correct replies.

Happy virtualizing! JP Please consider awarding points to helpful or correct replies.

View solution in original post

0 Kudos
3 Replies
AndreTheGiant
Immortal
Immortal

IMHO I usual follow the storage vendor's best practice for storage configuration and path management.

But note that (whitout 3rd part tools) also if you have 2 active connection, you can use only one (for each LUN).

Andre

Andre | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro
jpdicicco
Hot Shot
Hot Shot

Always go with the storage vendor's recommendation. The VMware recommendation is based on generic array characteristics (controller failover methods, ALUA-capable, etc). The storage vendor's recommendation is based on their performance and compatibility testing. You may want to consider their recommendations carefully, however, to be sure that each point is what you want.

With the 8000, I have run with Round-Robin. It is the most robust pathing option available to you from a performance and failover standpoint, and can provide the most even performance across the ports on the storage controller.

While I have not done specific testing/validation, the last time I looked at the docs, the HP configuration recommended that you setup each I/O to switch ports in the configuration. This adds load to the ESX host, performing the switch across ports, but HP claims their testing showed that this was the optimal configuration. This was the only setting that I questioned in their recommendation.

If you haven't already, make sure you download the HP doc on configuring ESX and EVA arrays. It has several settings that you should configure in addition to the path policy, as well as some scripts to help with making the changes.



Happy virtualizing!

JP

Please consider awarding points to helpful or correct replies.

Happy virtualizing! JP Please consider awarding points to helpful or correct replies.

View solution in original post

0 Kudos
RobBuxton
Enthusiast
Enthusiast

Thanks for the responses.

Kind of what I was I was thinking, nice to have it confirmed.

As you've noted the Vmware docs seem to be more generic, the HP one is quite specific in it's recommendations.

Nice to know others have done this.

0 Kudos