Subject: [netatalk-admins] Multi-zone routing
From: Hans Gerwitz (hans@phobia.com)
Date: Wed Mar 04 1998 - 16:23:24 EST
Please read patiently, I don't want to leave anything out:
I've just put into production a FreeBSD 2.2.5 box (Dell Dimension 486)
with netatalk 1.4b2 (no patches). This machine has three network cards
(3Com 509b), one to our 'backbone' and the others to subnets. TCP/IP
gateway functions are working flawlessly.
I've commented out the two lines in the startup script for afpd and papd,
as I am only routing on this machine.
atalkd.conf is as follows:
ep0 -phase 2 -net 20 -addr 20.10 -zone "slu" -zone "ACAD" -zone "SLU MED"
ep1 -seed -phase 2 -net 2557 -addr 2557.1 -zone "RH_LAB"
ep2 -seed -phase 2 -net 2444 -addr 2444.1 -zone "MATH_CS"
Note that the first interface, ep0, is the backbone card and is
auto-configuring wonderfully.
The problem: AppleTalk routing appears to work flawlessly from the
subnets. A machine in zone RH_LAB, for example, has no trouble seeing
items in MATH_CS or any other zone.
>From outside these subnets, though, NBP registration is flakey for the
zones defined for the subnets (RH_LAB and MATH_CS). Selecting one of
these zones from any machine outside the subnets may or may not list
devices contained within. Note that the zones are always available, just
not their contents. Leaving the chooser open, one can watch the devices
occasionally appear, only to later disappear. Leaving machines
side-by-side (both physically and network-wise) on the chooser, you will
see only one see into the zones, while the other still cannot.
Once a connection is established with a device, however, it is reliable.
Said another way: NBP registration for seeded networks does not reliably
broadcast to the non-seed backbone, but is reliable on that and other
seeded nets.
Finally, some possibly relevant information. I am getting frequent
syslog errors from atalkd that read:
"rhrouter atalkd[160]: as_timer sendto: Network is unreachable"
These come in spurts of about 6 every 20 minutes or so.
As I write this, I just received a number of these messages from atalkd:
"rhrouter atalkd[160]: nbp brrq sendto 22921.0: No buffer space available"
There were many of these, each for a different network number (all to the
.0 broadcast), but not one for every network number on our net. I should
note that our AppleTalk network here is a mess, with many, many Novell
servers that have been improperly set up to seed the backbone, and to
define zones, even though they do not have subnets (some of these had
subnets that have been removed, most were simply copied from existing
misconfigured servers from admin that don't know any better.) The router
I've described, though, is the only one experiencing this problem, and
the Novell-based one it replaced did not experience this, using the same
network numbers in the same location.
Thanks in advance for any help you can offer.
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Hans Gerwitz gerwitzh@slu.edu
Academic Computing Development, Information Technology Services
Saint Louis University <http://www.slu.edu/>
This archive was generated by hypermail 2b28 : Sat Dec 18 1999 - 16:31:19 EST