vSphere HA not cover failures on vSwitch?

[fernandoquinino]

Hi everyone,

I'm studying vSphere HA and reading how vSphere HA detect host failures. Before I write my question here i'm going to describe in few works what I was studied about how High Availability works on vSphere. Correct me if i'm wrong or miss some details, please.

vSphere slave hosts send heartbeat over management network to master host. When master host stops to receiving these heartbeats from a slave host, it start to look for heartbeat signal inside datastore before declaring the host to have failed. If slave host not issuing heartbeats it is decreed to have failed and then the host`s virtual machines are restarted on alternate hosts. In case when slave host is exchanging heartbeats with a datastore, the master host assumes that it is in a network partition or network isolated and so can leave virtual machine on or power off and on in other host.

I read this and i really didn't saw any mechanism to detect and trigger vMotion in kind of failures with vSwitch, it's a reason i sharing my thinking here.

I imaging what happens in big environments with a many vSwitches used for virtual machines and have many networks paths to send heartbeat. In this scenario of failure nothing happens because master host remains receiving heartbeats. So, my virtual machines using this vSwitch haven't any network connectivity and they still on in a host with network failure. Why i din't move these virtual machines to other host with network connectivity?

I hope i'm wrong or have other way to do something in this scenario, anyone have ideas?

Tks.

[rickardnobel]

fernandoquinino wrote:

In this scenario of failure nothing happens because master host remains receiving heartbeats. So, my virtual machines using this vSwitch haven't any network connectivity and they still on in a host with network failure.

First, the actual vSwitches only exists in ESXi memory and should not be able to failure by themself. If you mean all incoming physical NICs fail into a vSwitch then the vSwitch will still "live", but only be able to communicate with other VMs on the same host and vSwitch.

If I understand your general question correct then you are right in that HA only detects host failures, in the sense that ESXi is either down (or lost access on the Management network and/or the storage network). However, there is nothing in HA that can detect if all paths to a vSwitch with only VMs have failed. So if the Management network is up, but all links to your VM vSwitch is down - then no HA action will take place.

[fernandoquinino]

Hi Rickard,

Exactaly when i said about failures on vSwitch it was meant all networks adapters associated with that had failures. I know which vSphere HA only detects these host failures but i want to know if have other way to provide HA in that situation.

Tks.