Subject: Re: problem mounting two servers
From: charlie wood (charlie@id-inc.com)
Date: Mon Sep 10 2001 - 15:30:55 EDT
Excellent. Thank you for this. Yes, this turned out to be the problem. For
anyone else interested in the solution, here's what I did:
First, I checked the hexadecimal host ID on both servers with the command:
hostid
...they were identical, which indicated that I was on to the problem.
I then corrected the default /etc/hosts file for each server to read:
192.168.x.x hostname.domain.org hostnameAlias
127.0.0.1 localhost
...correcting the localhost entry to read as above, and moving it down one
level, putting the correct IP address/hostname entry in the first level. I
then rebooted (because rebooting is fun and profitable for the whole family!
;-) ).
Worked like a charm.
Apparently, atalkd uses the Unix system's hostid output instead of the real
IP address (obtainable from the /etc/hosts file) to identify the server. In
my particular case, because both hosts were essentially "localhost", with
the loopback address creating identical hexadecimal "hostnames", atalkd
assumed that when I was connected to one server, I was connected to both.
thanks again!
</charlie>
_____________________________
Charlie Wood
IT Manager
ID, Incorporated
1620 SW Taylor St Suite 100 Portland, OR 97205
{v} 503-223-7737 {f} 503-223-2719
{e} charlie@id-inc.com {w} http://www.id-inc.com
--> From: Thomas Schierle <ts@visual-s.de> > Date: Mon, 10 Sep 2001 21:04:26 +0200 > To: Netatalk-Admins <netatalk-admins@umich.edu> > Subject: Re: problem mounting two servers > Resent-From: netatalk-admins@umich.edu > Resent-Date: Mon, 10 Sep 2001 15:10:16 -0400 (EDT) > > The problem has been discussed more than a year ago, I believe. > Unfortunately I didn't find the thread in my mail archives. > > If memory doesn't fail, the solution was to change the order of appearance > of the localhost entry and the servers real IP within the hosts file. > Real IP has to be the first entry. > > Background (from memory): a netatalk server creates some kind of > unique ID at (its first?) startup, and that ID is based on its IP. > If that ID is based on the localhost IP, it of cause isn't unique. > > -Thomas > > On 2001-09-10 17:44 +0200, James Litwin wrote: > >> Ya , I have seen it before ..... not to sure what caused the problem ... only >> way that I could solve the problem was to rebuild the server from scratch >> .... >> then everything worked ........ >> >> charlie wood wrote: >> >>> I've got two separate servers, each have unique network.node addresses. For >>> illustrative purposes, we'll call one Bob and the other Sally (names changed >>> to protect the innocent). Whenever one of Bob's volumes are mounted on the >>> client desktop, and then an attempt is made by that client to connect to >>> Sally, via the Chooser, instead of Sally's login dialog, the "already >>> connected to "Bob"" dialog comes up. Selecting "OK" takes the user to Bob's >>> volumes list. The only way to actually connect to Sally, under these >>> circumstances, is to either unmount all of Bob's volumes, or make the >>> connection to Sally via afp (clicking the IP Address button in the Chooser >>> window, typing in the IP address, etc.). >>> [ ... ] > -- > Thomas Schierle, Munich, Germany > > PGP key [DSS/DH] 0xA23CDA1D available at various public key servers > >
This archive was generated by hypermail 2b28 : Sun Oct 14 2001 - 03:04:51 EDT