Wireless Access

My mobility master (8.2.2.4 VM) has started logging this sort of thing, pretty much constantly. I'm not sure what it means.

routing_switch[7121]: <310318> <7121> <ERRS> <arubamm-name IP> |ofc-routing-switch| max host limit(current:3200, max:3200) reached, not learning new host (mac:client-mac)

What is the subnet size of the network(s) that your MM is connected to?

The MM is on a /23. The MACs listed in the log are all WiFi clients.

the message is coming because openflow has "seen" more than 3200 hosts across all vlans attached to the controller. You can run "show openflow-controller hosts" to get some idea of where those IPs are coming from.

You can consider to exclude unneeded vlans from openflow (see bind-vlan in the openflow profile), or disable it if you're not using it (airgroup, ucm, sdwan) etc.

Thanks JGoff, that's helpful information. Clearly need to do more reading on how this works. Airgroups not currently enabled, though we will want to switch that on. The MM doesn't see any of these hosts on its vlan, so this is the MDs reporting hosts to the MM via openflow. All the openflow config is default.

Can you point me towards any documentation that explains the 3200 host limit? I'm not sure what that's referring to. Given that for airgroups we will have many more hosts than that.

hi Matthew, step one is find out what is clogging up the OFA hosts - are they valid hosts (end users) or some other randomly conencted vlan that is unrelated to wifi (e.g. campus lan network inadvertantly trunked ot the controller, mgmt network etc) ?

I dont know if the limit is documented - I cant find it mentioned anywhere

regarding what happens if the number of Airgroup clients is > 3200, let me ask around internally.

That's a good point, I'm in the early stages of migrating from AOS 6.5 to 8 and the WiFi client vlans are tagged across the 6.5 local controllers and the cluster MDs, so the MDs will see MAC addresses on the wire that are not associated with the cluster.

Choosing two recent MACs from the log shows that one is associated with the OS 8 cluster, the other is not. That could explain this I guess.

In the absence of any more info, I'm going to have to open a TAC call about this. The 3200 host limit for openflow can't be right, especially considering airgroups supports 100,000 clients.... 

hi, yes, please do & sorry for the inconveneince, the person whom I chased on this left his position and was not able to look into it before he did so. thanks.

I am seeing this same error on our system.  Did you ever find a resolution?

I did. In my case it was because the virtual mobility master VM wasn't sized correctly. The size of the VMM(50, 1k, 5k, 10k) is defined by how many vCPUs you give the VM. Ours was at the default install spec with 3 vCPUs so even though it was licensed as 5k I think table sizes are smaller.

Thanks!  I have increased our CPU count and RAM to match what the sizing guide suggests and hopefully the messages in the log file don't come back :)