Shared Volume between 2 non-clustered nodes

[Karl Crowson]

I'm trying to create a SAN-based shared volume between 2 non-clustered nodes. Is it possible with HP-UX 11.11? One node will write files to the shared volume. The other will read those files from the shared volume and deletes them. Thank you in advance for any advices/pointers on how I can create such a shared volume.

[Patrick Wallek]

About the only way I would even consider doing something like this is with NFS.

Trying to mount a VG on 2 servers at once is very dangerous and you have the potential for filesystem corruption if both are trying to run operations on the same file at the same time.

[A. Clay Stephenson]

It's trivially easy to do this; simply import the VG on both nodes and then mount them read-write on both nodes. It'll work marvelously for at least a few tens of seconds. The problem is that neither machine has a clue what is going on in the other's buffer cache. Even the fallback mode of mounting read-write on one node and read-only on the other suffers from the same problem. The only safe way to hard mount cooked filesystems on multiple hosts is for all of them to be read-only. You could use raw/io on multiple node because now there is no buffer cache but even in this case there must be some mechanism to coordinate the writes. If you want read-write on both nodes using cooked files the ONLY method that works safely is that one of the hosts must use NFS.

You might look at veritas (www.veritas.com); they have a distributed filesystem that might be of interest to you.

[Steven E. Protter]

So long as only one tries to use it at a time, fine.

NFS or Samba on top of NFS is the only solution here.

Full SG as noted does not allow for simultaneous sharing.

Next.

SEP

[Victor_101]

HI guys ,
I have a similar question,I have a pair of rp5470s which are hooked up to an EVA3000 (Veritas DFS not supported) and also hooked up via the same Brocade switch of the SAN to a Engenio Storagetek DS with SATA drives which we installed as cheap & fast backup solution. Now were adding an rp2470 to the SAN environment which is running our test instance. The plan is, in case of disaster , i run my test instance from the backup on the SATA drives. The trick is how to share the volume between the 2 servers, the direct solution that i deduced from the above comments is to create volume on the rp5470 node and export the volume to the rp2470 and make the volume on the rp2470 as RW until i do have an actual disaster.
What do u think??

thanks
vic

[Sยภเl Kย๓คг]

Hi Karl,

I would recommend you to go for the NFS or SAMBA mount and purge those files which ever needs to be cleaned. As Clay said, u can even allocated the same LUNs to both servers and import the VG in the second server with a map file taken from first server(with VGID). But going for this may create a possible filesystem corruption, we have tried it here. It worked for some time, but went into hung state after some time.

Hi victor, We would request you to rise another thread, since two questions by two different autoers will create lots of confusions. What you can do in ur case, is to give reference to this particular thread in ur thread if u find both of ur problems are simliar.

Regards,
Sunil