I am struggling setting up a NIM master to serve hosts on a network that is not directly accessible through the "master"-s default interface. I now that I need to set up network definitions and maybe NIM routes, but got confused.
The master has 2 interfaces:
en0 holding the address which is belonging to its hostname, therefore the AIX default gateway is using that interface (let's call this Network-master1)
en1 holding IP address from another network. Hosts from this network are enabled to communicate with hosts from my target network, but that communication needs a static route to be set up in the OS - as the default gw routes in a wrong direction. (Let's cal this Network-master2)
And there is another network, holding the to-be-clients - Let's call this Network-client
Every object is served from the master. If I understand correctly, I only need NIM routes is some objects are served from different nodes, thus I would not need this. Is that right?
I tried to set up a second install interface for the master, but as it is part of Network-master2 and not Network-client, I was unable to make NIM to set bootp server address to its second interface IP. It always put "sa=IP.ON.FIRST.INTERFACE" and therefore the client is unable to get its boot package.
There are no error messages to examine. Every command is run with "OK" exit status, but the config created by the commands is not correct. Clearly the problem is that I am doing it the wrong way and I wanted to get some advices on the correct way.
I would like to make NIM use install interface "if2" but it is still usind "if1", thus configuring bootptab with incorrect server address.
First off: The NIM boot process relies on bootp/DHCP as well as tftp. The route from your NIM server to the client(s) have to let these protocols pass and they have to be capable to act as "bootp relay agents". If these preconditions are not met you can stop right here because it will never work.
On your NIM master you have to set up a "NIM-Network" first, before you create the clients. Basically it is a definition that clients in the IP-network X will be serviced over your local IP interface Y by Gateway Z. You can do that easily on the command line but - i suppose you are not quite the NIM-expert - i suggest you use SMIT: NIM -> Administrative Tasks -> NIM-Networks (or similar - i have no AIX system at hand now and am quoting from memory). You do not need to state routing or so, that is all done quite normally via the IP stacks means (routing table, ...).
You can list the networks you have so far with
lsnim -t network
By the way: it pays off to do that for every subnet you want to serve with your NIM-server. There is also the option of entering a short description of the network you are creating. Do that, as it helps you keeping track of the NIM configuration.
In your case, where you haven't really started to do something, you are right. Still, you should be aware of the various logs in the path /var/adm/ras as well as the nim -o showlog command, which can be used to display certain logs. Both sources can be quite useful in determining why (and if) a certain NIM command went wrong. Notice that the general NIM log ( /var/adm/ras/nimlog ) is not a text file - use i.e.
Well, almost. I have 2 interfaces on NIM master, connecting to 2 different networks and I want to install into a third network which is accessible trough interface2.
--Trifo
------ Post updated at 07:36 AM ------
Well, there is NO failed operation. Though, I know that there would be. I see the problem arising before the error would be logged.
There is an interface - if1 - belonging to the first interface of the master. There is a second interface - if2 - defined to the second Ethernet interface of the master. Both interfaces have their matching NIM network definitions. There is a third network definition as well. This third network can be reached through en2 - there is an OS static routing entry for that.
My problem is the following: how to tell NIM that I want to use "if2" when installing client into the third network. How to achieve that NIM would configure bootptab with the IP of en2 as server address (sa=x.x.x.x).
All other ports and traffics are enabled to travel between the two networks.
(and, well, I am not a real NIM expert, but I am fond of command line)
I'd rather not share internal network informations here fully open to the internet. I also contacted IBM support personnel and no luck there either.
I will keep on experimenting, but for now I had to have a solution, so I installed my hosts out of mksysb images from the VIO server.
Even though I will need NIM config later.
I have got the solution eventually: there were 2 routing entry in my network (ent) object. routing1 was the default, which were added when the ent object was created. routing2 were added later by me. I had to remove routing1 by issuing