# Basic transit problem in two node setup

**URL:** <https://discourse.haproxy.org/t/basic-transit-problem-in-two-node-setup/9586>\
**Category:** Help!\
**Created:** [March 6, 2024, 1:05pm UTC](https://discourse.haproxy.org/t/basic-transit-problem-in-two-node-setup/9586 "2024-03-06T13:05:26Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![m0535](https://avatars.discourse-cdn.com/v4/letter/m/e68b1a/32.png) [@m0535](https://discourse.haproxy.org/u/m0535)\
**Post date:** [March 6, 2024, 1:05pm UTC](https://discourse.haproxy.org/t/basic-transit-problem-in-two-node-setup/9586/1 "2024-03-06T13:05:27Z")

</div>

Hello Community  
I have a problem, the traffic will no pass the HAProxy:  
The configuration is based on two HAProxy’s with two nics of each.

A dump shows me input traffic but no output traffic.

```auto
frontend fe1
    bind 10.10.11.1:80, interface eth1
    mode tcp
    default_backend be1
backend be1
    mode tcp
    balance roundrobin
    server s1:80 10.10.12.1:80 check
    server s2:80 10.10.12.2:80 check backup

```

the virtual ip’s are deployed via pacemaker ipaddr2 between two HAProxy nodes.

Backend adresses  
HAProxy-1 : eth0 : 10.10.10.1/24  
HAProxy-2 : eth0 : 10.10.10.2/24  
defaulroute looks to eth0

Frontend adresses  
HAProxy-1 : eth1 : 10.10.11.1/24  
HAProxy-2 : eth1 : 10.10.11.2/24

As a local firewall i use nftables without nat. Drops will be logged, there are no drops.

Is there any nat or explicit routing needed if i want to hide the be-pool servers from the clients.  
And the pool servers have to speak with the backend ips on eth0.

So the traffic comes like this : Client \<10.10.11.0/24\> eth1 eth0 \<10.10.10.0/24\> pool servers

Is here anyone to help me to fix my problem. And yes a part of the problems can be are not explicit HAProxy configuration. I’m not sure…

sysctl.conf  
net.ipv4.ip\_nonlocal\_bind = 1  
net.ipv4.ip\_forward=1  
net.ipv4.conf.all.log\_martians = 0  
net.ipv4.conf.default.log\_martians = 0

Cheers Dave