Wireless Access

I have an existing 6.5 network with two controllers. Master/Local setup.

Recently I purchased two HW Mobility Masters with the goal of migrating the network over at some point this summer.

I have very basic configurations on these MM boxes at the moment. Mostly IP, SNMP, etc. They are both running 8.3.0.7 code.

As soon as I connect these MMs to my production wireless network(same subnet my 6.5 controllers are on), all my APs on my 6.5 network drop from their controllers. They will eventually reboot and reconnect, but will continue this pattern until I unplug them from the network.

Anyone have an idea of what might be happening?

Thanks.

I would triple check there's no duplicate IPs, that VRRPs are not using the same IDs/Passphrases, etc between your new MMs and your old 6.x infrastructure.

Passphrase for VRRP may be the same. I'll change that.

Thanks.

Actually, that change didn't help.

As soon as I enable the port on the MM, I get tons of these on my controller log, probably one for each AP:

Jun 5 16:15:13 KERNEL(Dana-343-Hall@141.222.114.51): [174860.352582] VAP device aruba100 created osifp: (d8d7f540) os_if: (da960000)

Thoughts?

Keep looking.

Ok - I did some testing, my users are probably irritated by the number of times they've lost wireless this morning.

I factory reset my mobility master.

After reboot, all I configured on it was an IP address on the management port and time settings.

As soon as I connect GE 0/0/0 to my production network I can no longer ping my 6.5 controllers.

I started messing with settings on the interface of the MM. As soon as I turned off spanning-tree, things got right. I noticed that the default priority for spanning-tree on these is 128. On a Cisco based network, that means it becomes the root bridge for that VLAN, since typically the root bridge in Cisco world has a priority of 24624. This would definitely explain the odd behavior on that network.

I then made the interface a trunk port and re-enabled spanning-tree, as I noticed that is how my 6.5 controllers are setup. Things were fine with me doing that as well. The reason I didn't initially make it a trunk port is that  controllers would need multiple VLANs (I have some roles that belong to vlans other than the default), and I don't believe there is any reason why a MM would.

So I have a couple of questions:

1) Is there any reason to have spanning-tree turned on for the MMs? (I don't plan on connecting multiple ports unless I port channel them)

2) Is there any reason that an MM would need multiple VLANs?

Thanks.

Good find, and is kind of a common thing to see sometimes. You core switches should always have the root bridge defined to ensure it's always the root (set BID as '1' or whatever), that way anything that pops up that uses a different BID of a lower value won't take over. 

1. IMHO no, but that's more a philosophical descision based on your network architecture.

2. Again, depends on use case. Some people have separate networks where they want the MM addressed on two L2s (*NOT* common but happens, maybe AirWave is on a separate NMS VLAN and controllers report to the MM on another, etc). Generally though you are correct, one VLAN is most common on most people's MMs.

I will add, though, since the MM is essentially a virtualized switch, we default to many of the same settings and support the same features (VLANs, STP, etc). So that's why it's there. The MM and MCs share a lot of common code base. SO there's lots of commonalities that lead to why those features are there.