HA Configuration on diskless Hosts

[dnwhittaker]

I have 2 hosts in a DRS cluster, and I would like to enable HA. I've configured the userworld swap on both hosts to be located at /vmfs/volumes/isos which is a folder located on NFS. When I enable HA on the cluster, one host is fine but the other errors out. I'm assuming that I can't point the userworld swap to the same location. How do I fix this?

Any help is greatly appreciated.

Don

[Dave_Mishchenko]

Have you tried it with a sub-folder?

[dnwhittaker]

Dave,

I've tried that too -- no success.

[dnwhittaker]

Dave,

I'm attaching a Word document showiing my setup and the error message I get on one of the Hosts. I've also included a screen shot of the Asvanced Settings for the ScratchConfig. Hope you or someone in the Community can help.

[dnwhittaker]

Do I need to create a sub-folder under vmfs/volumes/isos for each Host? When I was at Vmworld I was told I didn't, but when I'm looking at the ESX Server 3i Embedded Setup Guide page 30, it says, "The configured directory must be unique across hosts". I assume that means each host needs a unique folder.

[dominic7]

The userworldswap directory does have to be unique. We've been naming ours /vmfs/volumes/<some lun>/<servername-swap> so that each host has a unique swap location. I'm not 100% positive, but I was under the impression that the directory had to be on a VMFS volume, which makes your case with the NFS storage a bit interesting. Have you already opened an SR?

[dnwhittaker]

Dominick

I had this set up on a SAN LUN with a unique name for each host, but I still ran into HA issues. After a number of calls to tech support there still wasn't any resolution. So I am plugging away on my own.

Don

[dominic7]

We had the same issue where the last node to join an HA cluster would fail to configure, but like I said once we moved to unique folders for each server to have their own userworld swap area, the problem vanished. Again, I don't know how well this applies to NFS.

[bhgordy]

each 3.5i server must point to a unique userworld swap file.

ScratchConfig.ConfiguredScratchLocation

e.g.

server1. /vmfs/volumes/SAN1

server2 /vmfs/volumes/SAN2

reboot and your good.

fyi.. if you point both servers to the same userworld scratch location, you'll get a big red bang on the secondary server.

[vancod]

I have 6 new ESXi hosts.

I have tried pointing them to each to their own LUN (formated VMFS)

I have tried pointing them all to the SAME LUN

I have treid pointing them each to a directroy on a unique LUN

I have tried pointing them all to a directory on a single LUN

In all cases I've rebooted all hosts.

In all cases they continue to come back on line showing an error in HA config because userworld swap is not enabled.

?????

[Dave_Mishchenko]

Are you setting this prior to joing the cluster or after? What format are you using for the datastore path?

[vancod]

Thanks for the reply Dave. I seem to have resolved the issue.

I am unclear why using a LUN per host did not work, but I now have all 6 configured to use a single LUN with unique target directories. e.g.:

/vmfs/volumes/"LUN_NAME"/scratch_01, 02, 03, etc (where LUN_NAME is the sym link to the volume in question)

Once saved and rebooted it resolves the sym link to the (I believe) LUN UUID

I was doing this prior to joining them in the cluster

[sonofthor]

I know this is old thread, but they probably all need access to the same LUN so that they can all share access to the swap files in the event of vmotion or having to restart vms that were on a crashed node in the cluster etc...

So in that regards it makes sense. I'm pretty sure this is in the documentation somewhere actually.