nfs mount problem

[Claudio_17]

Hello,

I've a problem mounting a nfs fs.

If I do

mount itsap01:/ndm/GentranSource /users/ediadm/GentranSource

I have this error

nfs mount: get_fh: itsap01/ndm/GentranSource: server not responding : RPC: Timed out

itsap01 is a rilocable server in SG cluster.

If specify i.p of itsap01 dosen't work ; If i specify i.p. of itsap02 ( phisicall server ) it works .

Any suggest ?

Thanks

[Graham Cameron_1]

Looks like an address lookup error.
Does "nslookup itsap01" return a sensible answer ?
Likewise "nslookup i.p of itsap01 " ?
DO they match ?
Can you telnet/ping to either/both ?
-- Graham

[Mark Grant]

I am not sure from your post is itsap01 a package in the service guard cluster and itsap02 the real machine?

If this is the case then the only thing I can see that might be an issue, bearing in mind it works from itsap02, is routing. Try loging onto itsap01 and see if you can ping/traceroute/telnet back to the machine you are logging in to.

[Sanjay_6]

I have seen this problem when the nfs server is within the local network and the nfs client is behind load balancers / firewalls. When the nfs client sends requests to the nfs server using the relocatable ip address the reply that it gets is from the server fixed ip. That may be the reason the nfs client is refusing the nfs mount since it is expecting a reply from the relocatable ip address.

Hope this helps.

Regds

[Stuart Abramson_2]

The problem is that you are doing the export from the server instead of from the package.

Is the exportfs contained in the package control script, or do you run it after the package control script has brought up the package?

We have code in our MC/SG scripts to export filesystems. I think it came with the SAP for MC/SG tool kit. I'm not sure if it's part of standard MC/SG, but maybe others can help you there.

[Sanjay_6]

In my case when i had to nfs export a filesystem which was part of MC/SG cluster package, i added it maunally through the package control script. The export was done as part of the customer defined run commands and the unexport was done as part of the customer defined halt commands.

Hope this helps.

Regds

[Claudio_17]

Thanks all ,

ok Sanjay ; I believe you're right .. but what can i do ?

[Sanjay_6]

Claudio,

I wan't able to find any solution. We ended up using the server ip address to nfs mount the filesystem on the client. What we did was added the package control script to page the admin every time the package is bounced, This way the Admin is aware that he may have to take manual action to nfs mount the filesystem if the package bounced to the alt node. In case the package was just shutdown and restarted on the same server, the admin may not have to do anything since the nfs client would be looking for the mount and would automatically mount the nfs exported filesystem.

The problem we ran into was the NAL (Network Address translation) on the F5 load balancer side and that was the reason the nfs server reply was being rejected.

Hope this helps.

Regds
Sanjay