NFS issue, phantom, connecting but not?

[nyplnyc]

I'm trying to setup vmware HA with 4.1 and NFS.

The hosts are in an HA setup and I'm trying to present NFS to the first host.

Under configuration -> storage I added the NFS mount point. It shows up in the datastores list but it fails in recent tasks with a time-out error.

I refresh the datastore list and a second NFS mount point appears with (1) in the name. Every time I refresh I get another NFS mount point.

I tried to restart the service (/sbin/service nfs...) but part of that fails. Tried to stop nfs service and a bunch fail (maybe the stop worked?), try to start and it takes forever on starting NFS daemon. Status shows nfsd stopped after.

What am I doing wrong?

Also, is there a particular order of adding NFS mounts to a HA setup?

[avarcher]

Hi

Sorry, I dont have an answer here - I'm experiencing the same problem and wondered if you got a fix?

I'm not setting up HA, I'm just connecting to the NFS datastore.

Thanks.

[kac2]

there shouldn't be any issue. it sounds like you are doing the right things to me. just make sure that your hosts have root access to the NFS datastore and you have it on a seperate VLAN or physical network from the rest of your VI.

[avarcher]

Thanks.

It's the self same datastore we have been using since NFS came to ESX in v3.0, nothing has changed on it, been through all the ESX vers since without a problem and I now have 3 hosts on 4.1 and they are all showing this behaviour.

Pain in the butt.

[avarcher]

Thanks.

It's the self same datastore we have been using since NFS came to ESX in v3.0, nothing has changed on it, been through all the ESX vers since without a problem and I now have 3 hosts on 4.1 and they are all showing this behaviour.

Pain in the butt.

[nyplnyc]

Well, for me it was very frustrating - mainly because the error messages were just useless: "Operation timeout" or "invalid paramter specified".

After searching the boards I found one article that had the second error listed, but it was from a developer using the SDK.

Reading it carefully, I figured out what the issue was: An NFS/NAS cluster in which the host can be directed to different IP, in my case an Isilon.

Even though you can specify an FQDN when configuring the NFS mount point it seems like VMWare remembers the IP. So, when I configured it initially, it was all fine. When I hard code 1 IP it works fine.

I'm only testing this setup, so I haven't found the actual solution to this. I'm guessing there is some way to tell VMWare to not care that the NFS IP is changing. Don't know. There is also an ESX Best Practice Guide for Isilon which is on my to-read list.

[Mitchell_Orr · Accepted solution]

see http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=100443...

Your NFS server (in my case a NetApp FAS2040) may be registering in DNS with more than 1 IP address for the NFS name.

NSLOOKUP of the host name must only report 1 IP address for the name used in the NFS Datastore, or VCenter gets confused.

I'm still trying to figure out how to prevent the NetApp from creating more than one IP for the DNS entry, but I'm fairly sure that will resolve the problem.

Of course, you can simply put the IP address in within VCenter, but I prefer to use DNS names.

[nyplnyc]

Correct - this is exactly the problem. Not sure why or how, but ESX is looking for one particular IP even though it has a FQDN for main NAS cluster IP. Anyway, as this was a POC for us, I just set the IP and it worked fine. Not sure what the real way around this is - I imagine if folks have a NAS cluster, they want to use it.