[Mailman-Users] A cluster issue I didn't find in the faq, or on google

[Richard Nelson]

Mailman: 2.1.10
Python:  2.5.2
Debian:  Sid (unstable)
Linux:   2.6.25.3

I've got a small cluster of machines doing smtp/web/etc.  The storage is NFSv3,
moving to NFSv4 as soon as I get a few more Kerberos issues squared away.

To enable Mailman,  I symlinked mailman/archives and mailman/lists to a NFS
directory (thus keeping the queues local to each machine).

Most of this works just fine, any of the machines accept posts and even the
mail->(single machine for the nonce) news gateway is working.

Here's the odd part (making me think I just missed something obvious)

* on each machine, mailman/bin/list_lists works fine (shows the whole enchilada)
* on each machine http://<machine>/cgi-bin/(listinfo|admin)/<list> works fine

* on all but the original machine, http://<machine>/cgi-bin/(listinfo|admin)/
  says there are *no* public lists (when in fact there are several)
and
  on the original machine, http://<machine>/cgi-bin/(listinfo|admin)/
  shows the full list

There are no errors being generated (/var/log/mailman/error is emtpy),
no http errors, but cgi-bin/(listinfo|admin) think there are no directories !?!

My gut feel is that while the python scripts handle the symlink fine, it seems
the compiled cgi-bin somehow doesn't like the symlinks...  but that doesn't fly
because the original machine (which works) is using the same symlinks to the
same place...  The account running mailman is the same on all machines (ldap):
uid=38(list) gid=38(list) groups=38(list)

I'm stumped... any clues on what to look at next ?
-- 
Rick