Why do I have multiple paths to storage with 1 NIC?

[Wasisnt]

We are trying to get our setup to where there is only one active patch to a datastore by only setting 1 active NIC on a switch or port group. It works fine for certain datastores but not the others. Our storage is active/passive and the datastores on LUNs accessed through controller B show one path which is what we want and the datastores on LUNs accessed through controller A show 2 paths to the datastore which we dont want. Ive attached some screenshots.

[Wasisnt]

This is what we want it to show and could have sworn it was working the way we wanted earlier.

[foodandbikes]

Looks like your storage device has multiple IP addresses assigned to it. A path will be created to each IP.

Why do you want only a single path?

[Wasisnt]

It has 4 IPs assigned to it now for 4 iSCSI ports on the storage. What I want to know is why does it show 2 active paths when there is only 1 active NIC? On another datastore it only shows 1 active path for 1 active NIC.

We were having issues with multiple paths going to the same LUN over 2 different controllers and it was causing major problems. I always thought on an active\passive array that you dont want the same LUN accessed over 2 different controllers at the same time so we are trying to solve the problem while we work on a better network setup.

[kjb007 · Accepted solution]

Active paths and active NICs are not the same thing.  One active NIC on the server side, but multiple destinations, yields multiple paths through that single server NIC, but to multiple destination/storage NICs.

If you really want to limit the number of paths, then you will have to limit your destinations as well.

Additionally, on that screen where you can see/manage the available paths to the same storage, you can right-click and disable the path that you don't want to be usable.

-KjB

[Wasisnt]

Ok, that makes sense but how would you designate specific NICs to specific storage iSCSI ports? The bottom line is that we dont want the same LUN being accessed over both controllers at the same time on 2 different paths. Is this where port binding comes into play? I cant seem to find any real information on it.

[kjb007]

Also, although you have multiple active paths, only 1 is doing any I/O, based on the PSP (MRU) that you are using.  So, for that LUN, you will only use that single path, unless there is a failure.  Round Robin will allow you to use both at the same time, but that is not what you are configured to use.

As long as both active paths are going to the active controller, accessing through the 2 paths should not cause you any problems, but since you have only one NIC on the server side, you're probably not going to see much benefit either.

Hope that clears things up a bit.

-KjB

[kjb007]

Similar to fiber channel, since you are allowing multiple IPs available to the host, it sees them as potential paths.  If you limit those IPs, you limit the paths.  This can get to be more administratively taxing, but it's up to you how you want to control access.

Generally speaking, you should be able to limit which portals/IPs expose which LUNs, thereby limiting the paths that the server will see and can use for that specific LUN.

-KjB

[Wasisnt]

Is that done on the storage side? I guess my confusion is how you make sure access to a particular LUN is only done over one controller. On our storage, certain logical drives are assigned a preferred path (controller) and we need to make sure VMware doesnt use both controllers to access the same LUN. Or is it smart enough to do that on its own?

[kjb007]

VMware should be able to manage that on its own.  Looking at your active paths, those should all be pointing to the preferred controller, and the standby would be the 2nd controller.

-KjB

[Wasisnt]

Thats what I figured (and hoped). Thanks for the help!