Discussion:
problems with interface aliases.
(too old to reply)
Mike Mestnik
2006-10-20 21:41:16 UTC
Permalink
An old article, but an unanswered one.

http://mail-index.netbsd.org/tech-net/1997/08/24/0000.html

The fix for broadcast pkts is to pass the netmask and broadcast
options. I have not tested this and there is no way to set the ?unused?
NetworkID.

I have a bigger problem. I am setting up two systems with identical
IPs for HA reasons. Doing this prevents them from communication as
they once have, I'm ok with this.

Both systems are on different networks and both have different network
IDs. 123.89.89.89/24 and 123.24.24.24/24. After adding the aliases
123.24.24.24/32 and 123.89.89.89/32 respectfully. I add their alter egos
and put something in /etc/hosts so we can still use there names for
things like ssh, 123.89.89.90/32 and 123.24.24.25/32.

Darn, outgoing connections will not come from the alter ego as is
needed. I put these routes in place, but no go. Why is this not
working like a multi-homed setup? Should I get another ethernet
connection to the same network?

# 123.89.89.0/24 123.89.89.89
# default 123.89.89.1
123.89.89.90 123.89.89.1
123.24.24.25 123.89.89.90

# 123.24.24.0/24 123.24.24.24
# default 123.24.24.1
123.24.24.25 123.24.24.1
123.89.89.90 123.24.24.25

Vary wicked and vary not functional.
Should have the effect I have previously seen in multi-homed setups
where the originating IP is taken, indirectly, from the route used
(123.89.89.0/24 and 123.24.24.0/24).

I think I have it, I'm just messing things up some how. tcpdump is
not showing outbound on the one interface, there is no route to lo in
the path AFAICT. What should I be looking for?

4.11-RELEASE FreeBSD
--
/****************************************************************
* Mike Mestnik: Junior Admin 612-395-8932 *
* ***@visi.com VISI/Digital North *
****************************************************************/
Alt address: ***@spam.visi.com

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Bill Studenmund
2006-10-24 19:52:37 UTC
Permalink
Post by Mike Mestnik
An old article, but an unanswered one.
http://mail-index.netbsd.org/tech-net/1997/08/24/0000.html
The fix for broadcast pkts is to pass the netmask and broadcast
options. I have not tested this and there is no way to set the ?unused?
NetworkID.
I have a bigger problem. I am setting up two systems with identical
IPs for HA reasons. Doing this prevents them from communication as
they once have, I'm ok with this.
I'm sorry, why not use CARP for this? I thought HA was what it was for.

Take care,

Bill
David Young
2006-10-29 11:22:15 UTC
Permalink
On Fri, Oct 20, 2006 at 04:41:16PM -0500, Mike Mestnik wrote:
...snip snip...
Post by Mike Mestnik
Darn, outgoing connections will not come from the alter ego as is
needed. I put these routes in place, but no go. Why is this not
working like a multi-homed setup? Should I get another ethernet
connection to the same network?
...snip snip...
Post by Mike Mestnik
Should have the effect I have previously seen in multi-homed setups
where the originating IP is taken, indirectly, from the route used
(123.89.89.0/24 and 123.24.24.0/24).
Try adding/changing the route using route(8) with the -ifa option to
set the originating IP.

What do route -n get 123.89.89.0 and route -n get 123.24.24.0 say?

Dave
--
David Young OJC Technologies
***@ojctech.com Urbana, IL * (217) 278-3933

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Mike Mestnik
2006-10-30 15:47:44 UTC
Permalink
Post by Bill Studenmund
Post by Mike Mestnik
An old article, but an unanswered one.
http://mail-index.netbsd.org/tech-net/1997/08/24/0000.html
The fix for broadcast pkts is to pass the netmask and broadcast
options. I have not tested this and there is no way to set the ?unused?
NetworkID.
I have a bigger problem. I am setting up two systems with identical
IPs for HA reasons. Doing this prevents them from communication as
they once have, I'm ok with this.
I'm sorry, why not use CARP for this? I thought HA was what it was for.
These systems are not on the same network segment, they also provide
geographical redundancy.

Still I'm wondering if the CARP project has a workaround for this?
I'm not sure if the PF causes too many changes in the TCP stack or
interface drivers of the OS, where I'm seeing this problem.
Post by Bill Studenmund
Take care,
Bill
--
/****************************************************************
* Mike Mestnik: Junior Admin 612-395-8932 *
* ***@visi.com VISI/Digital North *
****************************************************************/
Alt address: ***@spam.visi.com

--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Loading...