Originally posted by: m.ayaz

Hi All,
We noticed that on one of my cluster nodes the GPFS filesystem mountpoint is not visible when checked through df command. while GPFS is running we can go in the GPFS file system. when we type command pwd it tells us that we are in GPFS file system but we can,t see files in GPFS file system.when we type df command GPFS file system is not visible but GPFS file is running.

The system runs OracleRAC10g RAC 2 node cluster on AIX5.3 with GPFS 2.3.

for reference please see attached screen shots for both GPFS nodes.

Regards
MA

Originally posted by: flodstrom

Is the /oradata file system the only GPFS file system? What are the other file systems, like the /app1, /app2 and /oracrs? Just normal jfs2 file systems?

What does "mmlsmount all -L" say?

Other than that, it's difficult to say without further information. It may also be the case that the specific node has problems and that you may need reboot it. However, before doing a reboot, is your GPFS set-up with tie breaker disks?

Either way I would advice to contact IBM support about this.

Originally posted by: m.ayaz

Dear flodstrom,
Thanks for your response.

• /oradata & /oracrs are the GPFS file systems.

• mmlsmount all -L result is

root@dowpdb2 [/]> mmlsmount all -L
ksh: mmlsmount: not found.
root@dowpdb2 [/]>

• Yes tie breaker diks are available in GPFS.

Let me tell you one thing our GPFS server (node 2) was down for 3 month due to some reasons. After 3 months when we turn ON our servers we are facing the problem as mentioned above.

Originally posted by: flodstrom

Just realised that mmlsmount is a GPFS 3.x command, sorry about that.

Did you contact IBM support about this?

Other than that.

On node 2, when you run df do you get the prompt back at once or does it remain "stuck" for a while?

Do you know if the working node was updated (patched) while node 2 was offline for 3 months? It's important that all GPFS nodes are on the same minimum level. Run "mmlsconfig" and check what level of GPFS is installed on both nodes (lslpp -l "gpfs.*").

Does the private network between the two nodes still work? Does a ping from node 1 to dowpdb2_interconnect work and obviously the opposite as well from node 2 to dowpdb1_interconnect. What's also important is to be able to freely login as root between the two machines without having to type in a password. To test that it's still working do a "ssh dowpdb2_interconnect" from node 1 and the opposite from node 2.