HA Testing - VSphere

[b4ndit]

Hello,

I recently deploy a Vmware Virtual Infrastructure.

We are using 5 hp blade server and the san was msa2000fc with Brocade SAN sw interconnect 8/24.

All 5 server had 3 LUN, each size is 2047GB

Then, I like to try VMWAre HA features. VMWare HA will take action if an ESX server contains VM(s) server in it was suddenly unavailable (unplanned downtime). VM(s) will restart at any available host in the cluster.

The virtual environment are like this..

We had 5 ESX server, ESX1-ESX5

All ESX server had 3 Datastore

Datastore1 contains 3 VM(s) and operating at ESX1.

ESX1 only had 1 FC port (using mezzanine FC HBA single port)

Thing I would like to test, if I disable HBA port at ESX1 (disabled from SAN sw management), will those 3 VMs restarted at another ESX server?

Any comment are appreciated..

Thanks.

[marcelo_soares]

If you have VM monitoring enabled on this HA cluster, yes, the VMs will stop and will be migrated to another ESX, where they will power on again.

Marcelo Soares

VMWare Certified Professional 310/410

Virtualization Tech Master

Globant Argentina

Consider awarding points for "helpful" and/or "correct" answers.

[weinstein5]

Disabling the HBA port by default will not cause an HA event since the ESX host is still running - what you need to do is enable the per VM monitoring feature of HA - when the VM< crashes because it can not reac its storage that is when HA will restar the VM on another host - if you want to test the host level HA I would use the ilo care for the blade center and powere odwn the blade - and if HA is configured cor4rectly the VMs will restart on the other blades -

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

[admin · Accepted solution]

No, HA does not currently respond to HBA or storage array/nework failures.

Elisha

[admin]

Not all vms will crash when they lose access to storage - it depends on the OS. So vm monitoring may not get triggered. And even if it did vm monitoring currently only restarts the vm on the same host - it doesn't try moving the vm to another host.

Elisha

[MattG]

If you have redundant HBA ports/paths to storage, then you should incur no downtime.

To test HA you can:

• Disable your Service Console NICs in the blade management and it cause an HA isolation event.

• Hard power off a blade in the blade management. The VMs to should be restarted on another host.

-MattG

If you find this information useful, please award points for "correct" or "helpful".

[b4ndit]

I've tried this kind of lab yesterday..

the result is VM not responding, seems like VM stuck, they are not moving, HA not triggered..I've waited for 30 min..nothing happens..

Cluster setting (HA,DRS) VM monitoring was used.

So, to avoid this kind of situation, I am using alarm, so if ESX lost its storage, the host will turned off, then the HA will triggred..all VM jump to another ESX server in the cluster..

happy ending

Thanks for your comment folks..

[Vishy1]

I have tested HA failover in ESX 3.5 using the following method, move a test VM to isolation host

Set your HA cluster setting to Host Isolation response: Power off VM

In ESX4 cluster setting set Admission Control to enable: Do not power on VMs.....

Login to SC on isolation host, execute this command

esxcfg-vswif -D ( drop heart beat for HA and isolate the host)

wait for VM to restart on available host

esxcfg-vswif -E (enable heartbeat)

Ensure that you run this command from physical console otherwise you will lose ssh connectivity.

If you found this information useful, please consider awarding points for Correct or Helpful.