[LARTC] Backup Route

[Mike Fedyk mfedyk@matchmail.com]

<PRE>bert hubert wrote:
&gt;<i> 
</I>&gt;<i> On Thu, Nov 16, 2000 at 02:49:33PM -0800, Andrew wrote:
</I>&gt;<i> 
</I>&gt;<i> &gt; &gt;     I wan't to do a backup connection with the ADSL link.
</I>&gt;<i> &gt; &gt;     With what tools I can do this?
</I>&gt;<i> &gt; &gt;     ipchains, iptables, a third party tool?
</I>&gt;<i> &gt; &gt;
</I>&gt;<i> &gt; I'm in the middle of doing something very similar. I don't know how often they
</I>&gt;<i> &gt; update the archives on this list but the subject on the thread for what I'm
</I>&gt;<i> &gt; working on
</I>&gt;<i> 
</I>&gt;<i> Linux has a thing called 'dead gateway detection', which might work for you.
</I>&gt;<i> Just make two routes with a different metric, should work I think.
</I>&gt;<i> 
</I>&gt;<i> Existing sessions will die however because their source IP address is
</I>&gt;<i> unroutable.
</I>
I'm not sure that is working correctly.  I have two DSLs to the internet with
different IP addresses, speeds and ISPs.  

DSL1: 63.194.. 384Kb/sec X 1.5-6Mb/sec ADSL on eth1 via pacbell
DSL2: 64.63.. is 768Kb/sec SDSL on eth2 via lmki

Test: 
ping yahoo site through DSL1
unplug phone line from bridge on DSL1
ip route flush table cache
ping yahoo site --- No response
tracepath yahoo site  --- Still trying to use DSL1

That isn't working.

Here's my setup:
$ ip route
10.1.1.1 dev eth2  scope link  src 10.1.1.2
10.0.0.0/24 dev eth0  scope link  src 10.0.0.1
127.0.0.0/8 dev lo  scope link
default
        nexthop via 63.194.239.201  dev eth1 weight 99
        nexthop via 10.1.1.1  dev eth2 weight 1

10.1.1.1 is the internal ip of the dsl2 router.  Pretty nice, Pacbell should
start using these.
Note: the higher the weight, the more traffic sent through that route.
I want most of my local traffic to go through DSL1.  Anything requesting traffic
on DSL2 will get a response from DSL2, but very little traffic initiated by the
router machine will go through DSL2.

$ ip rule
0:      from all lookup local
32763:  from 10.0.0.0/24 lookup 10
32764:  from 63.194.239.202 lookup 10
32765:  from 10.1.1.2 lookup 20
32766:  from all lookup main
32767:  from all lookup default

$ ip route show table 10
63.194.239.201 dev eth1  scope link  src 63.194.239.202
10.0.0.0/24 dev eth0  scope link
default via 63.194.239.201 dev eth1

$ ip route show table 20
10.1.1.1 dev eth2  scope link  src 10.1.1.2
default via 10.1.1.1 dev eth2

I could write a script that would monitor the interfaces with ping and change
the default route after several failed attempts.  With the way ping works, to
account for several failed responses, you would have to wait about 60 seconds
for about three attempts.  I think 5 would be good, but that can be
configurable.

Does this sound good or is this already done better in another tool?  I wonder
if a routing daemon would do something like this.  Monitor with ping, but not
expect any routing messages from the monitored routers.  Anyone know?


</PRE>