mortis2600
Contributor
Contributor

High availability question

Jump to solution

I have a vm that I need to have high availability on. Host and VM monitoring is pretty simple, but I have a separate network that if it goes down (it's my wan) I need it to fail over. I'm not sure how to accomplish this.

Any ideas?

0 Kudos
1 Solution

Accepted Solutions
daphnissov
Immortal
Immortal

There's nothing built into vSphere that will proactively fail a VM over to another host in that situation. What sounds like needs to happen is you should design that WAN network with better redundancy and on your ESXi hosts as well (multiple switches, multiple uplinks, multiple and separate pNICs inside your hosts) to mitigate a situation where there are no SPOFs. This is just one of those things where there isn't automation built-in to compensate for infrastructure design decisions. The only other possibility I can think of, if you didn't want to rectify your networking from a physical layer, is to have some automation that sits out to the side and does periodic checks on that VM. If the checks fail (however you determine what constitutes a failure), then the automation can fail that VM over to another host. But even in such a situation, you could have issues where, again because of networking, the connection doesn't even come up once failed over because of something upstream.

View solution in original post

0 Kudos
6 Replies
daphnissov
Immortal
Immortal

What infrastructure do you have here (you didn't describe it) available to fail over on? Are we talking separate sites? Separate clusters under the same vCenter? What, exactly?

0 Kudos
mortis2600
Contributor
Contributor

Same site. The VM network is on its own physical switch. The wan is on a separate switch.

The WAN would come in and be connected to a switch. From there I would have an uplink going to each host. Those uplinks are om a WAN VSS attached to the VM.

The hosts have a VM network VSS and a management VSS. They are currently in a single cluster. VM in question would be acting as my SD-Router, and I want it to fail over if wan access is lost.

Honestly I'm not sure if that's possible without some serious work.

0 Kudos
daphnissov
Immortal
Immortal

But what would it failover to? Another host? If this VM has two NICs, one for VM network and the other for WAN, in what scenario would WAN be lost? If the host running that VM fails, you would use vSphere HA to have it restart on another host (which also has that same uplink pattern). Otherwise, if WAN connectivity is lost (which would have to be detectable through some external automation) I'm not clear what corrective action you're saying would need to take place.

0 Kudos
mortis2600
Contributor
Contributor

It would fail over to the second host.

Currently they both only have one nic each with 4 ports. One port is connected to switch the wan is coming into and is on its own VSS.

What scenarios would wan be lost? For example, the network link for the wan went down on a host. Bad port, cable or nic.

If the wan itself goes down, failing over wouldn't do much good. But I do want some redundancy for local network failure.

0 Kudos
daphnissov
Immortal
Immortal

There's nothing built into vSphere that will proactively fail a VM over to another host in that situation. What sounds like needs to happen is you should design that WAN network with better redundancy and on your ESXi hosts as well (multiple switches, multiple uplinks, multiple and separate pNICs inside your hosts) to mitigate a situation where there are no SPOFs. This is just one of those things where there isn't automation built-in to compensate for infrastructure design decisions. The only other possibility I can think of, if you didn't want to rectify your networking from a physical layer, is to have some automation that sits out to the side and does periodic checks on that VM. If the checks fail (however you determine what constitutes a failure), then the automation can fail that VM over to another host. But even in such a situation, you could have issues where, again because of networking, the connection doesn't even come up once failed over because of something upstream.

View solution in original post

0 Kudos
mortis2600
Contributor
Contributor

That's what I figured. But I wasn't sure. I appreciate your time listening to my question and answering.

0 Kudos