This post is more than 5 years old
2 Intern
•
227 Posts
0
3395
April 4th, 2014 12:00
Port Group Strategy
new to the VMAX world and am trying to understand what the best strategy is for dividing up the ports on the back of each director. i've come across several cases where we have hosts pointing at a group of ports going down the same director path. i'm looking to clean things up a bit but am hoping for some suggestions?
# of port groups to create?
# of ports to include in a given group?
port groups based on function? (Recoverpoint vs vsphere vs physical sql)
below is a quick overview of what I'm looking @ today:
Any feedback would be appreciated. as you can see I have room for improvement, just looking for a recommendation.


Quincy561
1.3K Posts
1
April 4th, 2014 12:00
Any given host device (LUN) should not be active on the same FA CPU (both ports 0 and 1). For maximum availability for just two paths, each one should be on a different director board. If more than two paths, the ports should be on different IO modules. E+F are one IO module, G+H are the second IO module. Since the IO module is a FRU (Field Replaceable Unit) you want to minimize the paths impacted when you replace hardware.
duhaas1
2 Intern
•
227 Posts
0
April 4th, 2014 12:00
great, thanks Quincy56 . question, appear unisphere for vmax recommend a min of four ports/port group? that accurate? or a it depends? not using powerpath of anything at the moment, just trying to lay things out a bit more efficiantly
Quincy561
1.3K Posts
1
April 4th, 2014 12:00
4 is a good starting point for most applications. Applications that don't require much performance can get by with 2. High demand apps may need more than 4.
Check out my best practices presentation. Some info on this subject.
Brainshark
Quincy561
1.3K Posts
0
April 4th, 2014 13:00
Yes, each director has 4 Front-End CPUs. I'm not sure the best way to answer the second part of your question, Hopefully others will chime in.
duhaas1
2 Intern
•
227 Posts
0
April 4th, 2014 13:00
Also, to confirm from this comment, are you saying that each director has four CPU, one backing each group of ports? Also, whats the best way to introduce new ports to port groups? just zone host to access on the backend and introduce the ports and remove other from port group?
duhaas1
2 Intern
•
227 Posts
0
April 4th, 2014 13:00
Thanks again Quincy56 GO BREWERS!
cincystorage
2 Intern
•
467 Posts
0
April 4th, 2014 23:00
You can simply zone the hosts the the new ports, add the ports into the port group, verify the devices are showing up on the host via the new ports, then remove the old ports from the port group, and remove the zoning.You just need to be very careful and make sure you double check that everything is showing up.
I notice you said you aren't using any multi patching? ESX has native multi patching built in. Without multi-pathing of some kind having any port beyond one is basically useless. What you end up with is a host that sees the same lun 4 times, but It wont understand they are the same device.
I highly suggest that as part of this clean up you're doing, and not after it or separate, you make sure of multi patching options available to you.
dynamox
9 Legend
•
20.4K Posts
0
April 5th, 2014 05:00
how do you remove FAs from a port group, that has devices mapped to it ? You think -remove -unmap will be sufficient or you have to make those devices write disabled on those specific FAs first ?
cincystorage
2 Intern
•
467 Posts
0
April 5th, 2014 15:00
Yeah, a remove and unmap will work, but i'm not a fan of doing it while IO is active on that path - tho it should not be a problem with proper multipathing.... I just like to be cautious. I like to stop IO down the ports from the host level, the remove the FA from the port group. Then cleanup mpio/powerpath.