Well, first off, the network IDs have to be unique, i. e. "subnet_1", "subnet_2", etc.. I suppose that was only a typo on your part.
To answer your question: yes, you will have to create separate NIM-networks for this. No, you won't have to configure 3 IP-aliases for your master.
The reason for this is: NIM operates using NFS (for "normal" package installations), bootp (for mksysb-installs) and tftp. All these protocols have to be - obviously - allowed by all the firewalls you might have in between your NIM-server and the prospective clients. In addition the NIM-server needs the standard gateway to each network to be able to act as a "bootp relay agent".
bootp (see RFC951, RFC1542 and others) works with broadcasts from (in respect of IP-numbers) anonymous interfaces and such packets would be dropped at the nearest router. Therefore bootp would not work over network boundaries. a "bootp relay agent" is a router which will take such a packet and forward it to a certain bootp-server and the same way tunnel its answer back to the client, so that bootp works in routed internetworks too.
When you are in sms menu for client (IPL where you configure the IP, Master IP subnet and gateway), say for IP 10.30.1.21 you will use the gateway of that client ex:10.30.1.1 and NOT the NIM gateway ex:10.20.1.1 . Same goes with 10.40.1.X subnet.
Is this something that I need to check with our network team? We currently have 4 different subnets in our AIX lpars and they can talk to each other normally.
Thanks for your help Bakunin.
P.S. Just joined the forums again after years of rest. I'm still trying to get the hang of using it properly.
kaelu
---------- Post updated at 10:55 AM ---------- Previous update was at 10:52 AM ----------
Thanks ibmtech. Will remember this, I think I had issue regarding this when I'm trying to restore our VIO server via NIM.
In principle: yes. In practice the overwhelming majority of switches/routers do this (at least this is my experience) per default. I just wanted to give the rationale behind having to state the default gateway for each NIM network.
Notice that "default gateway" here means the gateway used to access the client network: suppose you have a NIM-server with IP A out of network NetA and a client with the IP B out of network NetB. The NIM network you have to create will be NetB and its "default gateway" will be the router IP in NetB, not some interface on the NIM server. For example (suppose x.1 is always the router port for network x):
Thanks Bakunin. That helps a lot especially the detailed examples. I can't wait to build my new NIM server and do all of this. I have my design already and just waiting for it to implement.
Thanks for your hospitality as well and really glad to be back.