HA and DRS error

[Siew]

I setup a ESX server lab. I create a cluster call MSP, inside the cluster have 2 host they are server1.msp.com and server2.msp.com. and just have a VM (Linux Red Hat) inside the cluster.

The VMware HA options:

Linux Red Hat, Restart priority: high, Isolation Respones: Leave Powered on.

1st the VM power on at server1, after that i unplug the network connection from the server1. So from the server1 status is "not responding" shown from VC, and then the VM restart at server2 (meaning that HA is successful).

Then i wait for a while, and then connected back the network cable into the server1. VC is detected server1 is up, but for the VM become running from both host (server1 and server2). From the "VC ->MSP cluster->Virtual Machines" keep changing the VM running on host, it show VM run on server1, and then show VM run on server2, and then show VM run on server1, ... keep changing.

Can anyone tell and guide me to solve this problem.

[SyverDude]

Are they using shared storage? In a shared storage environment, the lock will be taken away from the old server and given to the new server.

\- Jon

[conyards]

Split Brain....

Because you have a two node Cluster when node one becomes isolated from the network, node two believes it's isolated too... You see by default they only check with each other to see if they are isolated...

Two ways around this... Install a third node... or look at utilising the advanced feature setting Das.IsolationAddress and point this at a network device thats up 24/7, thus giving the ESX host two network reference points to accertain if it has become isolated or not.

N.B. the advanced fetures tabs should only be used in a no production test environment or ideally in conjunction with VMware support...

Simon

[Siew]

Yes SyverDude, The VM is storage in nas, and both of the ESX server is map into the nas.

[Siew]

Conyards, can i set the Das.IsolationAddress and point to the Virtual Center? Because at the moment i don't have enough PC to do this.

[ZMkenzie]

Just set a pingable ip as a gateway for your vmware service consoles (the virtualcenter ip will be fine). Esx will check first other esx servers and then it will check the service console gateway, if no one is reachable then it will go in isolation mode. In this way no split brain is possible since virtualcenter acts as a third server.

Hope this helps.

[conyards]

The ESX servers gateway will be ideal for the das.isoalationaddress setting.

Simon

[Siew]

I have another question. if i add Das.IsolationAddress point to 10.0.0.191 (Virtual Center), and then i unplug the network cable from the ESX server1. Now the ESX server1 is totally isolated from the network, the ESX server1 can't ping to gateway, Virtual Center, and ESX server2 also.

For this situation, the VM will power on at another server which is ESX server2 because the HA feature. And Then because of the Isolation Response option is set to "Leave powered on" so the VM still remain on at ESX server1.

At this point, when I connected back the network cable into the ESX server1. What is the action Virtual Center will going to do? Because when I unplug the network cable, the VM is power up at ESX server2 but the VM is still remain up at ESX server1. And then when I connected back the network cable, It become that VM is running on two different ESX server.

Is Das.IsolationAddress able to solve this problem?

Thanks

[ZMkenzie · Accepted solution]

You should not set "keep machine powered on" if you think that another esx host will power on that machine. Having the same machine running on 2 hosts is a really bad idea even if the hosts can't see each other. This is why, in isolation mode, esx does a shut down of all the vms.