[comp.protocols.appletalk] Split-addressing with Fastpaths & atalkad

dswt@stl.stc.co.uk (Stewart Tansley) (06/12/91)

------------------------------------------------------------------------
Anyone else out there ever tried to used split addressing mode on their
Fastpaths (option 9 in K-STAR v8)? We're trying to reconfigure from a
simpler existing setup using one subnet. Any experience/help would be
appreciated...
------------------------------------------------------------------------

We have a simple setup: 3 Fastpath 4s handling 3 appletalk zones, bridging
to a backbone ethernet. Fastpath configuration is (or would be!) done
using an atalkad daemon running on a central Sun. 

Originally everything went along swimmingly with the Fastpaths & their
associated Mac virtual IP addreses in the same subnet. Now though, we have
to split this subnet into 3 new separate subnets, and place the Fastpaths
on the main net, using the 'split addressing' mode.

On setting things up in the new configuration, we can get one Fastpath to
run apparently fine (although associated CAP v5 hosts don't seem to be able
to do NBP stuff correctly -- getzones & atlook fails), but on connecting
up a second fastpath, we had a massive broadcast storm from the latter,
resulting in mega net problems. 

Temporarily we have gone over to local configuration of the other 2
Fastpaths, and they seem to 'mostly' work in split addressing mode; the 
single daemon-controlled Fastpath remains stable. CAP seems to provide
AUFS and LWSRV services ok, but atlook and getzones fail. We need to move 
to a position that allows daemon-control of these two other Fastpaths.

Questions:
- Has anyone got a copy of their atalkatab and atalk.local files that they
  could send me, where you are using a similar split-addressing setup?
- Any problems I should know about in this setup -- eg. why would we get
  such a broadcast storm? Does split addressing *work* using atalkad?
- To experiment, I may wish to setup a second atalkad -- can I run this
  while the existing one is running (it's control is out of my hands), but
  running different atalkatab files?
- Why me? :-)

Thanks for any assistance! :-)

===========================================================================
Stewart Tansley     | BNR Europe Limited (was STL...) |  'Be cool, or be
                    | London Rd, Harlow, CM17 9NA, UK |    cast out...'
dswt@stl.stc.co.uk  | +44 279 429531 x2763            | Subdivisions, Rush
===========================================================================
  'You know how that rabbit feels -- going under your speeding wheels...'
===========================================================================

morgan@JESSICA.STANFORD.EDU (RL "Bob" Morgan) (06/18/91)

> Does split addressing *work* using atalkad?

I haven't tried it myself, but certainly there isn't any way for atalkad to
deliver both an IP address for the Kbox and a different one for its client
range: there just isn't any place to put this info.  But remember that atalkad
does two things:

1) it delivers local config info to each Kbox (eg, number of clients, LocalTalk
net number and zone name, etc)

2) it delivers a routing and zone table to every box in its internet.

In general, #1 isn't crucial because you can configure all this info by hand
with FastPath Manager (FPM).  But if you want to have Kboxes form an AppleTalk
internet across IP-only paths, there isn't any substitute for the routing info
provided by function #2.  On the Kbox at least (dunno about other products),
you can't just type this info in.

It could be that you could configure the split addresses by hand using FPM and
still use atalkad to deliver the other local info to the box.  I've observed
that K-STAR is pretty good at using hand-configured info first where it exists,
and atalkad info for any other parameters.

Another approach is to hand-configure everything with the FPM, and just use
atalkad to deliver routing info.  I think it should work to just set up an
atalkad like:

24.9    E       158.36.19.40      Stanford
53.26   E       158.36.58.43      Twilight
95.3    E       158.36.171.10     Danger
73.8    E       158.36.22.40      Zilla

where the IP addresses are those of your Kboxes, and the net numbers and zone
names those of their respective LocalTalks.  Ie, replace the "K" lines that
you're no longer using with E lines that just provide routing/zone info to the
rest of the net.  It's worth a try.

Here's another approach.  I get the impression that you're doing this because
you've run out of IP address space on your original subnet.  What we do a lot
of (though it's controversial) is to use more than one subnet on a single
Ethernet (it's almost always Kboxes that make us want to do this).  Our cisco
routers make this easy to do.  This gives you N*254 (assuming you're using
8-bit host parts) IP addresses on the Ethernet; sorta like Phase 2, huh.  You
do have to make sure that boxes on the same Ethernet with different IP subnet
numbers still agree on AppleTalk net numbers and zone names.

> I may wish to setup a second atalkad -- can I run this
> while the existing one is running (it's control is out of my hands), but
> running different atalkatab files?

Sure.  It's not unreasonable to run two Kboxes with similar but non-identical
atalkatabs (just make sure they don't contradict each other).  If I have two
atalkatabs, A and B, users who are on nets defined only in A will only be able
to communicate with other nets defined in A (and similarly for B), while users
on nets defined in both A and B will be able to communicate with the union of
the two internets.  You can use this method to dig yourself into amazingly deep
procedural holes, if you wish 8^).  Of course, if two Kboxes using two
different atalkatabs share an Ethernet and do RTMP, they will exchange all
their routes and zones, with possibly amusing results.  

 - RL "Bob" Morgan
   Networking Systems
   Stanford


-------

johnw@group1.UUCP (John Wheeler) (06/21/91)

In article <4545@stl.stc.co.uk> "Stewart Tansley" <dswt@stl.stc.co.uk> writes:
>Questions:
>- Has anyone got a copy of their atalkatab and atalk.local files that they
>  could send me, where you are using a similar split-addressing setup?

    Sure:

    We have 8 subnets spread across 3 cities (San Francisco,
    Chicago, and New York) with FastPaths in each. The top 2
    subnets are unused, and the other 6 are divided as follows:
    1 -- SF regular TCP/IP stuff
    2 -- SF MacTCP on LocalTalk using split addressing
    3 -- Chicago regular TCP/IP stuff
    4 -- Chicago MacTCP on LocalTalk using split addressing
    5 -- New York regular TCP/IP stuff
    6 -- New York MacTCP on LocalTalk using split addressing

    Our CAP/AUFS host (we only have one set up right now) is
    running atalkad to configure the FastPaths.


Our atalk.local file:

#
# mynet        mynode        myzone 
5              1             IPTalk

# bridgenet    bridgenode    bridgeIP
5              3             192.101.31.3


Our atalkatab:

# the IPTalk networks:
0.5     N0  192.101.31.1     IPTalk       #SF      IPTalk
1.5     N0  192.101.31.65    CBOE         #Chicago IPTalk
2.5     N0  192.101.31.159   New_York     #NY      IPTalk

# the EtherTalk networks:
0.2     E   192.101.31.3     Ethernet    #SF      Ethertalk
1.2     E   192.101.31.67    CBOE        #Chicago Ethertalk
2.2     E   192.101.31.131   New_York    #NY      Ethertalk


# the FastPaths:

# Madonna:
# Mon May 20 11:36:03 PDT 1991
#----------------------------------------------------------------------------
0.10   KC    192.101.31.3    LocalTalk      #madonna, localtalk
    I192.101.31.31      L0                  #ipbroad ipname
    L0                L0                    #ipdebug ipfile  
    L0 L0 L0 L0     S0.2  S200              #ipother atnetet ddprangestart
    LX0        S16    S8                    #flags ipstatic ipdynamic
    S0.10     S0.5                          #atneta atnete

## Prince (Chicago):
## Mon May 20 11:43:49 PDT 1991
##----------------------------------------------------------------------------
1.10   KC    192.101.31.67   CBOE           #prince, localtalk
    I192.101.31.95      L0                  #ipbroad ipname
    L0                L0                    #ipdebug ipfile  
    L0 L0 L0 L0     S1.2   S200             #ipother atnetet ddprangestart
    LX0        S16    S8                    #flags ipstatic ipdynamic
    S1.10     S1.5                          #atneta atnete
#
#
# Sting (NY):
# Mon May 20 11:41:50 PDT 1991
#----------------------------------------------------------------------------
2.10   KC    192.101.31.131   New_York      #sting, localtalk
    I192.101.31.159     L0                  #ipbroad ipname
    L0                L0                    #ipdebug ipfile  
    L0 L0 L0 L0    S2.2  S200               #ipother atnetet ddprangestart
    LX0        S16    S8                    #flags ipstatic ipdynamic
    S2.10     S2.5                          #atneta atnete
#
#

>- Any problems I should know about in this setup -- eg. why would we get
>  such a broadcast storm? Does split addressing *work* using atalkad?

    Make sure you also have option 17 - transition mode - set. Unless it is,
    a "storm" of FwdReq packets seems to get generated instead of being 
    converted to lookups within the FP. We had this problem, too.

    Also, if you don'y have it, get K-STAR 8.1.1 from Shiva. It's available
    all over for downloading (CompuServe, I believe, other public places, etc.,
    they also have a BBS). It allows subnet masks to be entered in 
    'user parameter 3'.

>- To experiment, I may wish to setup a second atalkad -- can I run this
>  while the existing one is running (it's control is out of my hands), but
>  running different atalkatab files?

    I wouldn't try that....I have enough trouble keeping ONE running.

>- Why me? :-)

    Because you're willing to attempt the unproven...
>
>Thanks for any assistance! :-)
-- 
________  John Wheeler - Informix-4GL / SQL * Unix * Broadcast Audio Production
 \rrrr/   Group One Ltd.
   \/     San Francisco