LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Everything on the same subnet

To: linux-virtualserver@xxxxxxxxxxxx
Subject: Everything on the same subnet
From: Gian Filippo Pinzari <pinzari@xxxxxxxxx>
Date: Thu, 11 Feb 1999 00:29:26 +0100
Hello everybody,

it is not a long time I joined the list. I'm very interested in this
trend.
This effort looks to be very well integrated in Linux kernel and,
therefore, does appear to be very promising.

OK, this is the problem. I'm presently playing with two machines
staying on the same netmask. They both are RH52 and some parts
of their disks are mirrored each other using a software named
mirrordir (http://www.obsidian.co.za/mirrordir/). The first machine
has IP 192.168.2.50, the second 192.168.2.51. They already implement
some HA and automatic failover functions and I wanted to add
Virtual Server support using IP tunneling and the local node feature.
Until now, without success.

I spent a long time trying to understand why the whole thing didn't
work.
I followed the documentation outlined in README.tunnel and tried
also to play with IP aliases in order to let one of the two to become
default gateway for the other. I tried with Telnet and FTP services
(using ip_masq_ftp.o). The director machine (i.e. the one where I issued

the ippfvsadm commands) responded to requests as expected,  but
the other one... The only result was the disk light turning ON at the
time a request was entered, so I thought packets could arrive but
could not get back. I made a lot of attempts setting different tunnels
and SUDDENLY everything worked well . I saved in a text file the
whole ifconfig and route configuration, rebooted, set up again the same
way but, for reasons I don't understand, it didn't work anymore.

The day after I looked at DejaNews and found a lot of people reporting
weird problems on tunneling, so I gave up (and added a line to my
TODOs).

Next step was to recompile the kernel (2.0.36+vs0.6) for port
forwarding and local node support. I expected to be successfull at
the first try, but, also in this case... Some packets were going at
the other end but they couldn't find the way back.

I tried either letting Machine A and Machine B on the same subnet,
either changing IP of Machine B to 111.111.111.50 , adding eth0:1 to
Machine A as 111.111.111.51, setting default route of B to A and
pointing ippfvsadm to services on the 111.x.x.x net (that is, if I
remember: ippfvs -A -t 192.168.2.50:23 -R 111.111.111.50:23 -w 1
and ippfvs -A -t 192.168.2.50:23 -R 111.111.111.51:23 -w 1).
Of course, ipfw was always set up as stated in VS documentation.
As usual, each IP could be ping-ed, but only the local node could
respond to requests. Probably I'm wrong setting the routes.

May you suggest me simple configurations for port-forwarding VS
and tunnelling VS that MUST work? I believe the problem is due to
the fact that both machines are on the same netmask. Is there any
limitation in using IP aliasing that I should know about?

Ciao, Gian Filippo.



<Prev in Thread] Current Thread [Next in Thread>