NFS mount of Linux fails when package shifted to failover node

[Taurian]

Hi Everyone,

I am having a peciliar problem. I have a two node cluster on HP-UX machines. A third machine having Linux RHEL 5 is mounting directories exported from the HP-UX servers.

I have entered filesystems to export in the hanfs.sh file on HP-UX primary and failover nodes. When package is on primary server, the Linux server mounts the NFS filesystems correctly. But when I shift the package to the failover node, the Linux server gives the following message:

mount: ahlprdsbdb:/sapmnt/PRD/profile failed, reason given by server: Permission denied
mount: ahlprdsbdb:/sapmnt/PRD/global failed, reason given by server: Permission denied
mount: ahlprdsbdb:/usr/sap/trans failed, reason given by server: Permission denied


NFS filesystems are exported exactly the same way on both SG nodes. I've verified spelling mistakes and all.

The /etc/hosts files from all three nodes have each others entries and can resolve hostnames correctly. What can be the issue? Please help!

Environment details:
Node1 (Primary SG node):
OS: HP-UX 11.23
SGeSAP: B.05.00
SG: 11.18
NFS Toolkit: A.11.23.07

Node2 (Failover SG node):
OS: HP-UX 11.23
SGeSAP: B.05.00
SG: 11.18
NFS Toolkit: A.11.23.07

Node3 (Application server, not part of cluster)
OS: Linux RHEL 5



Lemme know if any other information is required. Thanks.

[Taurian]

Just to update that when Node3 (Linux) mounts any other directory (e.g. /tmp) of Node2, it mounts without any error.

Only the directories that were mounted on Node1 previously are giving this error.

[Taurian]

Hey guys..... no replies????

[Peter Nikitka]

Hi Taurian,

many people do not work on Sundays ...

I would first compare the /etc/exports - Files of the two HP-servers.

mfG Peter

[Matti_Kurkela]

Can both SG nodes resolve the name of the Linux system correctly?

If the IP_address -> hostname resolution does not produce the correct result on the failover node, the failover node will think the Linux system isn't authorized to use the exported NFS filesystem.

MK

[wci]

Can u please share entries you have in your
hanfs.sh under exports ( XFS[0] and the rest ).
Also the output of "showmount -e" when the package is running on the failover node.
If you have not exported to everyone you can check that as well.
also would be a good idea if you can have all package IP related hostnames on your /etc/hosts file.

WCI

[Taurian]

Hi Guys,

Thanks for the replies.

The hanfs.sh entries from SG Node1 are as below:

XFS[0]="-o root=ahlprdap:ahlprddb:ahlprdsbdb:ahlap1:ahlap2,anon=0 /sapmnt/PRD"
XFS[1]="-o root=ahlprdap:ahlprddb:ahlprdsbdb:ahlap1:ahlap2,anon=0 /usr/sap/trans"



The hanfs.sh entries from SG Node2 are as below:

XFS[0]="-o root=ahlap1:ahlap2,anon=0 /usr/sap/trans"
XFS[1]="-o root=ahlap1:ahlap2,anon=0 /sapmnt/PRD"


Where,

ahlprddb = SG Node 1 (primary)
ahlprdap = SG Node 2 (secondary)
ahlap1 = Node 3 (Linux node)



I've verified both /etc/hosts files from Node 1 and Node 2, they have all the same entries. Node 2 can also correctly resolve Linux node's hostname from IP.

Rgds,
Taurian.

[Taurian]

Hi all,

I don't have the showmount command's output with me currently, but below is the exportfs output from SG Node 2:

ahlprdap: [/etc/cmcluster/PRD]# exportfs
/usr/sap/trans -root=ahlap1:ahlap2,anon=0
/sapmnt/PRD -root=ahlap1:ahlap2,anon=0


I've also noticed that when package fails over, the nfs filesystems on the Linux box stop responding for a few minutes but then everythings starts working. BUT after this, if I unmount the filesystem I cannot remount them (I get the same Permission denied error).

This is baffling. Please help!

[wci]

Hi,

Why do you use /etc/exports and hanfs both.
I assume you can use hanfs since it will take care of your NFS exports whithin the cluster.
It looks your hanfs is different on both nodes and you can consider making node2's script similar to node1.
Are you running any firewall on linux system like iptables?

WCI