HA without DRS

[stkristobal]

Hi!

I've got a cluster set up with 3 nodes; all licensed with vSphere 4 Advanced. This license does not include DRS; which I gather is needed in order for automated migration to work. At least it doesn't when I try to set a host in maintenance mode manually.

That's fine. But it doesn't migrate anything when a host fails either (I tried killing one to see what happened).

So the question is; do I need DRS in order for automatic failover to work? If so, what does HA do? Just power on when the host comes live again?

I haven't gone real deep into vmware before, so I'm kind of a noob on these matters, so bare with me

[FranckRookie]

Hi,

DRS is in charge of spreading the VM load over the hosts to balance the cluster. HA is used to restart a running VM on another host when its current host fails.

You do not need DRS to do failover. If your test did not work, you should have a configuration problem.

Did you install your VMs on a shared storage? Can you start your VMs on any host of the cluster? Is HA enabled on all hosts?

Good luck

Franck

[weinstein5 · Accepted solution]

Welcome to the Forums - DRS and HA wprk independent of each other - HA will restart the VMs on the remaining nodes of the cluster. When you shutdown the host did you make sure there were VMs running on the host you shut down? Did you confirm when the other host came up that the VMs were running on that host or on of the other hosts of the cluster? I concur with the other poster sounds like a configuration issue.

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

[AntonVZhbankov]

>So the question is; do I need DRS in order for automatic failover to work?

No.

>If so, what does HA do? Just power on when the host comes live again?

No, HA restarts VMs from failed host on remaining. But here are some rules:

1) All VMs must be on shared storage

2) There should be enough resources to restart VMs if admission control enabled. Read about slots here.

3) HA should not be disabled on cluster level for these VMs.

---

MCSA, MCTS Hyper-V, VCP 3/4, VMware vExpert

http://blog.vadmin.ru

[stkristobal]

Thanks for replying guys.

The VMs are on an iSCSI SAN, and I can vMotion manually perfectly fine between the three hosts, so I'm assuming the Kernel interface is OK and it's not a network issue. I can't really see that many config options for HA; and I didn't see any obvious reason to change the default settings.

The cluster is not yet in production; and only has three VMs, so resources is not an issue.

What config options come in to play here?

[VirtualV]

could you place you logs here. "/opt/lgtoaam512/log/rule/vmwareclustermanager" to verify what really happen when HA initaite.

[stkristobal]

Where are these logs? On the vCenter server or on the hosts? If so, which host?

[VirtualV]

log file on ESX host which you are taking offline to test you HA configuration.Also if possible paste your HA configuration screenshot here. /opt/vmware/aam

[stkristobal]

Hmmm... may have been a glitch there; redid the test after upgrading to 4.1; and it seems to be working now. They're migrated of the dead host and powered on again, as they should be.

Follow-up question:

Is there any function which would allow a server to be migrated instantly if a host fails, without it going down at all? I seem to remember there being one.

[stkristobal]

And also; I've set up the vCenter as a VM. How would this affect HA if the host holding the vCenter VM fails? Or is the actual HA a function between the hosts, not between vCenter and the hosts?

[FranckRookie]

Switching immediately to another host is Fault Tolerance feature. It maintains a second ghost VM, clone of the first one, onto a second host. In case the active VM fails, its clone is immediately activated.

HA feature is hosted by the ESXs, vCenter is not doing anything except configuration. Even if vCenter is stopped, HA is still fully operational.

Regards

Franck

[stkristobal]

Ah, thank you all! I'm wiser now