I am working on project that need shared access to the FS from 2 AIX HACMP nodes.
This FS is created on a SAN disk and the VG that includes this FS is active on node A.
What options do I have so that this FS can also be accessed from Node B ?
We don't have GPFS, so that is not a option. I can do NFS mount, but we tend to stay away from NFS due to mount issues.
I fear that there is no other way to achieve what you want. The problem is that the disk driver/file system level is simply not designed to allow for simultaneous access. You have to either have an application software on top (like a DB, which will coordinate parallel read/write processes to a common data set) or a specialized driver/protocol - like NFS. Either way you have only one system being responsible for data being read and written (on the FS level), not two.
I hope this helps (understanding, sadly, not solving the problem).
bakunin
Thanks for the response.
This is for a Message Broker project with the Multi Instance feature, eliminating the need for HACMP.
Required is that it needs a reliable shared FS, and redbooks are recommending NFS V4.
My dilemaa is if I get the FS allocated to 1 server (Node A) and NFS mount it to both the nodes.
My config is Node A (MQ/MB) Node B (MQ/MB). If Node A is NFS server can Node B also configured as NFS server to provide redundancy?
What happens when the NFS server (Node A) that goes down.?
I have tried to go thru the forums and the redbooks but I am not able to get a example of this .
These two nodes, now are they in an HACMP cluster or not? If so nothing easier than what you ask for. You put that NFS share into a shared VG on SAN and mount it eg. in /sharedFS_jfs2 on the home node where the VG is active. Through cluster NFS exports and statements in the RG definition you export it to both nodes. On both nodes then you mount homenode:/sharedFS_jfs2 to /sharedFS and have the share available on both sides with the same path. = NFS cross mounts.
Of course the service IP of $homenode needs to be part of the RG as well, so in case it fails over to nodeB that system will still be able to access the share. Having the VG with the same major number on both nodes eliminates the need for remounts after failover... The RG needs to varyon VGs and mount file systems before starting TCPIP related stuff like the exporting.