I've been asked why two active Networker servers cannot be clustered together so in the event of failover two Networker servers cannot run on the remaining node. As far as I am aware this is not possible but I don't have an answer yet why it should be impossible - our UNIX SA says that it should be possible to ensure the two logical nodes running on one physical host would have separate hostname, IP and port ranges so and the software can always be relocated to a different location for each hostname so why should this be.
Well, in my opinion it won't be possible because there 2 ports which will be common: 7937 and 7938 and those ports cannot be changed (the rest of service ports can be changed but not those two). That's why it won't be possible to run two instances of NW on one host.
Hmm, but I'm now thinking that if you will have a ability to bind NW to specific IP (not to listen on all IP addresses available on the machine) that you can run many instances of NW on one host (each attachedd to it's specific virtual address). But I have no idea how to do this.
I wasn't entirely bought by the claim the physical nodes would be able to run with identical port ranges on different IPs, but it could be possible to reassign ports - however I don't believe ports 7937 and 7938 can be reassigned
I can imagine (from the network point of view) that you can assign the "service" (whatever it is) to the specific IP address. But I have no idea (and never heard about that) how to make NW to bind itself to specific IP address (but if anybody knows that it will be magician trick).
Yes, that's possible as SA said, but this is not running application on the same logical host, but two logical hosts which then may share or have allocated resources on one physical machine. So binding 2 same applications to different IP and using different dbs is not possible unless you use logical partitions where each one runs as separate OS, but then this is seen to outside world (and applications running inside) as separate machines (I'm running server and storage node that way).
No it doesn't. I'm running it as NPAR under HPUX. These are seen as separated machines by the outside world. From physical point of view, all NPARs are the same physical box making them logical. From OS point of view and resources too, these are separate entities.
benzino1
2 Intern
•
244 Posts
724
0
Posted August 6th, 2009 08:00