vSphere HA agent on this host could not reach isolation address

[Gonzouk]

Hello,

We have 8 x 5.5u2 hosts and from time to time we get the warning vSphere HA agent on this host could not reach isolation address: 192.168.28.254

This only happens on the same 3 hosts at the same time. Never on the others and they are configured the same.

I can ping 192.168.28.254 and the hosts IP on 192.168.28.x during this time.

The host vis SSH can't ping it's gateway or trace to it:

~ # traceroute 192.168.28.254

traceroute: Warning: Multiple interfaces found; using 192.168.28.73 @ vmk2

traceroute to 192.168.28.254 (192.168.28.254), 30 hops max, 40 byte packets

1  *

Then after about 20 mins all 3 are fine again.  I'm trying to see if it happens every day around the same time, how can I check this?

This hosts go into Cisco switches and I can't see any errors or loops.

The problem I can see that may not be helping is the management network and the production vm guest network are the same.  Also the iSCSI netowrk is routable again not good, but this is how is has been for a long time it seems.

During this time no vm or users have any issues.

Any ideas where to start?

[cjscol]

How many vmkernel ports have you got enabled for Management?

How many vmnics are there connected to the vSwitch the Management vmkernel port(s) is connected to?

What load balancing is enabled on the Management vmkernel port(s).

Do you have multiple VLANs on the vSwitch the Management vmkernel port(s) use? If so are all the VLANs trunked on each of the physical switch ports the vmnics used on this vSwitch are connected to?

Are the physical switch ports the vmnics are connected to configured as an Etherchannel? If so you will need to have the load balancing on the vSwitch set to "Route based on IP hash".

[Gonzouk]

How many vmkernel ports have you got enabled for Management? 2 in the vSwitch0 shared withteh production VM network

How many vmnics are there connected to the vSwitch the Management vmkernel port(s) is connected to? 2

What load balancing is enabled on the Management vmkernel port(s)- Port ID

Do you have multiple VLANs on the vSwitch the Management vmkernel port(s) use? just 1 well it's set to vlan 0, but the etherchannel is set to a vlan at the other end

If so are all the VLANs trunked on each of the physical switch ports the vmnics used on this vSwitch are connected to? the are 6 vSwitches, iSCSI, vMotion, Trunk for varius vlans

Are the physical switch ports the vmnics are connected to configured as an Etherchannel? not all

If so you will need to have the load balancing on the vSwitch set to "Route based on IP hash" - all etherchannels are.

The production/management network (1 physical nics) go into 2 Cisco stacks (1 in each), I want a singel stack at some point, but it was how the VMware consultants set up a few years ago, so the management/vm prod VLAN is split into separate switches but in the same vlan.

As I mention the other 5 are setup like this and are fine.