Two installs of DCNM for SAN (for having a backup) possible?

[lpphiggp]

Our install of DCNM (just for SAN management, and unlicensed since our SAN is comparatively small) currently resides on a vm, whose volume is supplied from a datastore ultimately supplied by the SAN itself. It occurred to me that if something went really belly up with our array or SAN,  I wouldn't even be able to get to our DCNM to troubleshoot the issue, a bit of a catch-22.
Is it possible to have two installs of DCNM, and keep one as a sort of running "backup" as long as I kept it updated, without creating any kind of conflict?

The backup install would likely be put on a physical machine, but since we're short on resources it would also likely be a PowerEdge that is close to warranty expiration, so I don't want to rely on it as the primary.

[Manish Gogna]

The high availability and redundancy configuration for DCNM should be as per the following guidelines

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/mds9000/sw/7_0_x/configuration/guides/ha/DCNM/ha_dcnm/ha.html

HTH

Manish

[lpphiggp]

Thanks Manish, but I'm not asking about redundant supervisors or switches, just whether I can install DCNM on two different servers, in case one server crashes.

[GRANT GATHAGAN]

Since it's the unlicensed version, you can do it.

The issue is that the two installations will each have their own database.

If you are an experienced database person, there is probably a way to configure the two installations to share a database.

If that is possible, you run up against another issue: Which server should host the database?

If it's your current server, then a server or array problem will leave you without a working database.

If it's the old PE server that you don't trust, you'll lose the database if that physical server goes south.

You might be better off cloning the VM through your virtualization management to another datastore on a regular basis.

If there is any storage space on the host hardware that is independent of the array, maybe you could create a datastore for that purpose on the server itself.

If the VM is relatively small, and if its only purpose in life is to run DCNM, you could even shut it down and copy the VM's folder in the datastore to your local drive.

[lpphiggp]

Thanks, Grant, that's more along the lines of what I needed to know.

I figured two different servers would have their own database each, but then, don't they just pull their information from the switches anyway?  I only have a pair of MS9148s, nothing particularly complicated.  There are other switches (dell server chassis switches) but they're all in Access-gateway /NPIV mode, so not an actual part of the fabric.  DCNM is unlicensed, this is all I use it for.

Cloning sounds the like a good quick answer, actually, so long as it is stored on a local drive..(if it's on the datastore, then that too is using the SAN itself); that is the server's only purpose. It's probably larger than it needs to be (100GB hard drive, Windows Server 2012R2)

[GRANT GATHAGAN]

The DCNM does poll the switch upon initial installation, but recall that the DCNM is also the mechanism that you use to roll out changes to your SAN switches.

The concern is what would take place in the following scenario:

1) You create a 2nd DCNM and it polls the switch for the configuration.

2) You make changes via the first instance of DCNM, apply the changes to the switch and save its configuration.

3) You start up the second instance of DCNM again and it sees a conflict between its database and the switch's configuration.

I'd like to think that Cisco is going to include a mechanism where DCNM will ask you which is the correct setup; it's database or the switch's configuration.

Logic would say that the switch's configuration would take precedence over the database, but that is an unknown.

I can test that theory, but it'll take a week or so.

I have some new HBA's that I am going to be adding to my fabric.

I can clone the DCNM, make sure the clone operates properly, then add the HBA's to my current DCNM and see what the clone does when I run it after the addition.

I'll report back the results.

If you have any colleagues who are good at database work, I still think it would be feasible for the DCNM instances to share a database.

There may be some limitations to Cisco's implementation, however, particularly with the unlicensed version of DCNM.

[lpphiggp]

Ah, I see... that makes sense.  Thanks for your help!
I don't really have anyone good with databases around here that would have the time to sync them up, though that's an interesting idea.
If you get a chance to test your setup with the 2nd DCNM, I'd love to see the results, thanks again.