rapatel@athos.rutgers.edu ( Rakesh Patel) (12/22/87)
We have an unusual problem. Has anyone tried running CAP on a dual interfaced machine? We have two kboxes running KIP 0987, one on each of two subnets. The machine having the problem is connected to both the subnets. The host in question has addresses: 128.6.5.46 128.6.4.4 And the kboxes are: 128.6.4.240 128.6.5.108 The problem is that I can set up atalk.local as follows: # mynet mynode myzone 55.5 46 LCSR5 # bridgenet bridgenode bridgeIP 55.4 240 128.6.4.240 This entry makes CAP services accessible on the appletalk network hooked to the kbox on subnet 4. The CAP services are NOT accessible from the appletalk connected to the subnet 5 kbox ( i.e. not found under chooser). The next entry reverses the accessability: # mynet mynode myzone 55.4 4 LCSR4 # bridgenet bridgenode bridgeIP 55.5 108 128.6.5.108 In this case, the CAP services are only accessible from the appletalk on the subnet 5 kbox. CAP services are NOT accessible from anywhere when attempting to use the following: # mynet mynode myzone 55.4 4 LCSR4 # bridgenet bridgenode bridgeIP 55.4 240 128.6.4.240 or # mynet mynode myzone 55.5 46 LCSR5 # bridgenet bridgenode bridgeIP 55.5 108 128.6.5.108 Here is what our atalkatab looks like: 55.5 N1 128.6.5.0 LCSR5 #lcsr 55.4 N1 128.6.4.0 LCSR4 #lcsr 56.1 KC 128.6.4.240 CCISA1 #LCSR admin I128.6.4.255 I128.6.5.58 #ipbroad ipname I128.6.5.58 L0 #ipdebug ipfile L0 L0 L0 L0 S0 S0 #ipother unused unused L0 S0 S14 #flags ipstatic ipdynami S56.1 S55.4 "CCISA1" #atneta atnete zonea 56.2 KC 128.6.5.108 LCSRA1 #LCSR admin I128.6.5.255 I128.6.5.58 #ipbroad ipname I128.6.5.58 L0 #ipdebug ipfile L0 L0 L0 L0 S0 S0 #ipother unused unused L0 S0 S9 #flags ipstatic ipdynami S56.2 S55.5 "LCSRA1" #atneta atnete zonea The reason why we made all the zones different was to keep broadcasts at a minimum on our fairly busy ethernets. Anyone have any ideas as to why the strange behavior, or better yet, how to fix the problem? Atalkad is run on 128.6.5.58 if that matters. Note that all our other machines have worked fine. The only difference being that this one happens to be dual interfaced to the two ethernet segments that we use as shown above. Rakesh Patel.
cck@CUNIXC.COLUMBIA.EDU (Charlie C. Kim) (12/22/87)
The assumption here is that the two subnets are disjoint and connected by the particular machine in question. Remember, broadcasts, including subnet broadcasts, are not forwarded by a level-3 bridge (router). Since the N1 entries you have installed mean to use subnet broadcasts for NBP, you will indeed have difficulties. The reason all the other machines "worked" is that they probably communicated to the KIP boxes nearest. However, you will probably find that a machine on network 128.6.4 will not be able to "see" services on a machine (other than the KIP box) on network 128.6.5. There are two possible solutions - one I do not wish to describe because it involves undocumented (at the present time) features. The easy solution is to use the redirector (atalkrd). Exactly, how you will want to set it up depends on aspects of your particular network. Charlie C. Kim User Services Columbia University
cck@cunixc.columbia.edu (Charlie C. Kim) (12/23/87)
Well, people should probably ignore the analysis in the last message I sent about this subject. I keep forgetting that our subnet routers and configuration are not "normal". In particular, our subnet routers refuse to forward directed broadcasts on subnets... The "solution" part may or may not help out. Charlie C. Kim User Services Columbia University
hedrick@athos.rutgers.edu (Charles Hedrick) (12/23/87)
Just for the record, our problem turned out to be that our Sun 4's don't understand the 4.3-format broadcast address. We use a subnetted class B IP address. The machines for which I have source have all been modified to understand 128.6.4.255 as a broadcast address. I have been unable to get source for the Sun 4, so it still wants to see 128.6.4.0 as the subnetted broadcast address. The obvious change to the configuration file causes the gateways to use 128.6.4.0 for the local broadcast address, and everything works fine. It turns out that atis does not support the configuration we would prefer, which involves using different zones on different interfaces of a two-interface machine. Atis is only able to support one zone, and there can only be one copy of atis on a given machine. But that's no big deal for us. The reason we were able to talk to a Kbox on net 128.6.4 from our 128.6.5 interface but not from our 128.6.4 interface is that 128.6.4 and 128.6.5 are connected by a cisco router. The cisco router not only supports directed broadcasts, but it also lets us set a local broadcast address separately for each interface. It normalizes all broadcast addresses that it produces to that address. We use 255.255.255.255. So when the broadcast went through the router, the address got turned into all one's, which the Sun 4 understands. So we had the odd situation that local broadcasts didn't work, whereas broadcasts that went through the router did, because the router was normalizing the broadcast address to something that all of our machines understand.