[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

R: Re: iproute2 load balacing

hello, thank you for your response.

Configuring the multipath as I 
showed you, I need to split the traffic between the two wan. The 
traffic, however, does not come from the LAN, but is degeneratedby the 

Using this rule:

ip rule add fwmark 3 table 3
ip rule add 
fwmark 4 table 4

iptables-t mangle-A OUTPUT-p tcp - dport 25-j MARK - 
set-mark 3
iptables-t mangle-A OUTPUT-p tcp - dport 110-j MARK - set-
mark 4

but it also happens that packets coming out from the right 
interface but with the ip of the other WAN and viceversa

So I decided 
to use SNAT --to  ip[eth1] and SNAT --to ip[eth2] ,forcing a correct 

But ,I did not understand why, I must constantly flush the 
routing cache otherwise I lose packets and dont work.

Messaggio originale----
Da: pascal.mail@plouf.fr.eu.org
Data: 18-mag-
2012 11.30
A: "ninnig1@tin.it"<ninnig1@tin.it>
Cc: <debian-
Ogg: Re: iproute2 load balacing


ninnig1@tin.it a écrit :
> I have doubts about the management of the 
multipath with iproute2.
> In particular, do not understand how is 
the distrubution of traffic
> between two gateways.
> Use this 
> ip route add default scope global \
>    nexthop via dev eth1 weight 1 \
>    nexthop via dev ppp0 
weight 1
> Actually I see that some servers will reach from, others
> from10.64.64.64, but I do not understand the 
logic that uses this command.
> Can you help me understand?

multipath does not distribute traffic but routing cache entries.
routing cache entry is based on the source and destination addresses,

and is used to route all packets matching these until it expires.
means that all communications from A to B will use the same path

because they use the same routing cache entry. The distribution

algorithm is a simple weighted round robin and does not take the amount

of traffic on each path into account.

The efficiency as a load 
balancing increases with the number of routing
cache entries, i.e. the 
number of sources and/or destinations.


Reply to: