Wireless Access

I'm not able to get Aruba OS 8 MM running on our VMware infrastructure. Wondered if anyone knows how to get around this problem.

I've used the OVA file, which all looks fine. The VM comes up and goes through the initial setup. After a reboot it either reboots itself with a datapath timeout error, or core dumps. If I can manage to log in before it reboots or dumps, I can't show the config.

I had the same result testing with Virtualbox. 

I've built a virtualbox machine and installed the ISO. This appears to work. I've tried reinstalling the VMware machine from the ISO but no joy, just core dumps or reboots.

Our VMware is fully patched and up to date.

What ESX version and patch version? Also can you post a screenshot of the VM settings for your MM? 

Also try disabling Network adapter 1, 3, and 4, leaving only network adapter 2 connected. And then do you have promiscuous mode enabled on your vSwitch/port group that the MM is connected to?

We went with the settings specified in the OVA file. I've noticed that's below the minimum required for the MM to run. So it now has 6 CPU, 8GB ram, 2 x 8GB discs and three network interfaces. I haven't enabled promiscuous mode, but wouldn't expect that to cause a core dump even if it doesn't work.

VMware ESXi, 6.0.0, 4600944

I've just reinstalled using the new settings, will see if that helps.

Sadly not, it still core dumps. :(

I don't have access to change VM settings so will try disabling network interfaces tomorrow when the virtualization team are back in.

I wouldn't think so either, but I've seen weirder when the datapath is messed up. Otherwise you can open a TAC case, but screenshots of the VMSettings would help, specifically the network settings and after disabling network adapter 1.

Settings were changed to just leave NIC 2 connected. Hasn't helped.

Here are the settings.. unfortunately the html5 vsphere client doesn't lend itself to taking screenshots of the settings.

I haven't reserved CPU and RAM for this VM as I'm just at the 'take a look at it' stage. But again, this shouldn't kill it. I'm wondering if it doesn't like the network interface type.

I've just tried creating the network interfaces with a type of E1000 rather than VMXnet and they're not recognised.

Thing is, we started by importing the applicance file so that really should just work. 

Be happy to work through what the problem is.

VMXnet should be the type that is used. It has / supports more advanced network feature sets on the VMNet side. You SHOULD be able to change it back and it work fine.

Thanks Jerrod, we started out with VMXnet. Initially just let the appliance file do its thing. I've tried increasingly weird things to get this VM to work but I've had no luck at all.

We don't believe there's anything odd about our vmware infrastructure and don't have these issues with VM appliances from anywhere else. I'd really like to get to the bottom of this, but can't open a TAC case because I have none of the required details such as serial number etc. Might give it a go with n/a in the webform and see what happens.

So might be a steep ask, but if you can, post screenshots of your vmnet config (network settings on VM as far as network adapters), your vSwitch and port group settings, as well as what your uplinks are. I would also make everything access VLAN 1 JUST to make sure you're flat to start, then if you make an L2 change and it breaks you know there's an L2 issue in the middle (between VM and physical network). 

I've requested that from our VM team and will see what they come up with.

Hi, we're not entirely sure what you're asking us to do here....

We don't use vlan1 anywhere on our network. The MM vm is told it has an access port, which it does on the appropriate vlan for our infrastructure.

What I'm not clear on is irrespective of whether the VM can successfully communicate with the rest of the world, it shouldn't fail to interract with the virtual NIC, which it appears to, and it shouldn't kernel panic.

I'm trying to arrange a screen share session with TAC, along with someone from our VM team. It's starting to feel like the OS8 VMs are not compatible with VMware 6,  but I don't have enough information to be certain of that.

Get the screenshare session setup with TAC. 8.0 MM and VMC work fine on ESX 6 (it's what I'm running on now). 

Ok, thanks that's good to know. Got a session booked for Wednesday. 

So.... It looks like the issue is related to the vNetwork Distributed Switch. If we configure the VM with a vNetwork standard switch then we have no problem. We can kill the MM VM by switching it to our infrastructure standard of the vDS, almost immediately kernel panics. 

My guess is it's something to do with the driver for the paravirtualised interface. Could also be a VMware bug, but we're not in a position to do any further testing of this.

For now I've got things running using an unstable dev server, which isn't ideal, but at least we can start playing with OS 8, and the issue we've found is nicely reproducable (at least on our servers).

Some feedback to close this off.... 

The Mobility Master VM requires an MTU of 1500, and can't cope with jumbo frames.

Our vmware environment's distributed switch has a global MTU of 9000. Even though the port group we used for the MM was limited to an MTU of 1500 the MM driver doesn't like it and kernel panics. This has been acknowledged by TAC and passed on to development.