Virgin Media Community

[kpanchev]

Hi everybody,

I have recently had an interesting experience and was wondering if someone has had same or similar issues.

Basically my setup is this:

SuperHub 3 in modem mode

Own router running on a virtual machine (dd-wrt) on a Vmware esxi host with a dedicated NIC and a static MAC address on the VM

This has been running without any issues for the last several years, but recently I decided to tweak and streamline my environment a bit further. As I have two identical esxi hosts running in a DRS cluster, it was only logical to ensure that the router VM could be automatically migrated to any of the hosts. So to implement this feature, I created a separate VLAN on my Netgear managed switch with three ports only, one connected to the SuperHub and two connected to dedicated NICs on each host.

So far so good, the setup is working, the router VM could be migrated to any host without any noticeable service interruption.

The issue I had happened during a migration which coincided with a heavy traffic to the shared storage for the hosts, which lead to the router VM stalling. During the revival process i did a WAN DHCP address release and attempted a renewal, which proved to be unsuccessful.

So starting to troubleshoot, I noticed that if the VM is connected directly to the SuperHub, the DHCP process will work. Then if the link is transferred through the switch, it will continue working, also it will renew the IP address, but if I attempt release and then renew, it will not work.

Further investigation showed that I had jumbo frames enabled across the switch, which also has been the case for the modem VLAN. The hosts NICs have been set to MTU 1500, also I believe this is the setup of the SuperHub (no way to check or change)

Currently I have set the MTU for the VLAN to 1500, but have not had a chance to test the setup and see if this was the problem, but just wondering if someone here is familiar with the specifics or has experienced such a problem before? It will save me some wireshark packet sniffing in order to try and figure out what the issue was...

[gary_dexter]

What third-party router are you using?

The hub will only hand out one DHCP address to one attached device.

[Adduxi]

The Hub will not hot swap MAC addresses for WAN IP. If your VM router has two different MAC addresses when moving hosts, then the Hub needs to be off when doing so. It’s just the way it is, I mean it’s a get you going device, never made for stuff like migrating VM routers on ESX hosts!!  😀  Maybe a fixed MAC address for the VM across the hosts may work? 

[Z92]

When the hub is in modem mode it is intended to be connected directly to one device only, not a switch, so your configuration is not supported.

Also, changing the MAC address of the attached device sometimes (but not always) requires a hub restart. Since a MAC change typically means equipment change, this is not normally a problem.

Personally, I have the VM Hub directly attached to a HW firewall, which is then connected to a switch. As soon as you start introducing VLAN's on the HUB side of things you are guaranteed to have problems as it's meant as a simple residential router and certain not business grade.

[kpanchev]

Guys you probably missed my explicit note that I have a static MAC address set on the router VM... and yes, it is not a third-party router, it is a VM (if you don't know what the abbreviation means, please do not reply to the post)

By the way, this post is not for the general responders who fill in their post counts with pointless comments, I need advanced specialists here, and also people who READ and UNDERSTAND the post before posting a random vanilla reply...

[gary_dexter]

And you expect people to reply and help when you make comments like that?

[Adduxi]

---

@kpanchev wrote:

<snip> By the way, this post is not for the general responders who fill in their post counts with pointless comments, I need advanced specialists here, and also people who READ and UNDERSTAND the post before posting a random vanilla reply...

---

Okay, I'm out then. 

[kpanchev]

---

@gary_dexter wrote:
And you expect people to reply and help when you make comments like that?

---

I am expecting the right people to respond, I don't have either the time nor nerves to read through comments from people who only think they understand stuff. If you feel insulted, please don't respond.

If there is no one here that could help, then I will probably discover the answer myself, after some extensive testing and troubleshooting, I was just wondering if someone could save me some time. So if you can't do this, at least don't waste any additional time.

[kpanchev]

Well, as I suspected, the issue was the Jumbo frames. Now the MTU is back to 1500, the setup works without any issues.

So thanks everyone, I can confirm that I have a working virtualised environment with automatic failover of the router VM. If anyone is interested in the details, I would be happy to provide more info.

[legacy1]

The MAC on each NIC would have to be the same for the failover to take over the switch would need to be setup so the so traffic from one NIC can't see the other and incoming the switch mirrors to both NIC's

In this setup both routers can be active and as long as you don't go over VM new DHCP server discover/request profile which looks to be about two for ONE WAN IP on one MAC it will work.