[CentOS] CentOS 8: two external network adapters, two ISPs - routing problems

Fri Jul 30 14:27:13 UTC 2021
Konstantin Boyandin <lists at boyandin.info>

Hello!

Given: a CentOS 8-powered computer with three network adapters.

eth0, eth2: external, connected to two different ISPs
eth1: faces home network (intranet)

The task: allow accessing certain internal services from either ISP. 
There are several services, I only mention SSH below.

In the configs below:
IP1: external IP at first ISP (ISP1), assigned to eth0
Gateway1: IP of gateway provided by ISP1
Network1,Netmask1: related to IP1
IP2: external IP at second ISP (ISP2), assigned to eth2
Gateway2: IP of gateway provided by ISP2
Network2,Netmask2: related to IP2
LocalSSHIP: IP in intranet (eth1) where SSH server is running

Current configs follow. Routing tables:

echo "200 isp1" >> /etc/iproute2/rt_tables
echo "201 isp2" >> /etc/iproute2/rt_tables

Routing policies:
/etc/sysconfig/network-scripts/route-eth0

Network1 dev eth0 src IP1 table isp1
default via Gateway1 dev eth0 table isp1

/etc/sysconfig/network-scripts/route-eth2

Network2 dev eth2 src IP2 table isp2
default via Gateway2 dev eth2 table isp2

Routing rules:
/etc/sysconfig/network-scripts/rule-eth0

from IP1/32 table isp1

/etc/sysconfig/network-scripts/rule-eth2

from IP2/32 table isp2

iptables snippets. External traffic forwarded to local SSH server from 
both interfaces:

iptables -A PREROUTING -t nat -i eth0 -p tcp -d IP1 --dport 22 -j DNAT 
--to LocalSSHIP:22
iptables -A PREROUTING -t nat -i eth2 -p tcp -d IP2 --dport 22 -j DNAT 
--to LocalSSHIP:22
iptables -A FORWARD -p tcp -d LocalSSHIP --dport 22 -j ACCEPT

eth0 is default gateway:
$ ip route

default via Gateway1 dev eth0 proto static metric 100
default via Gateway2 dev eth2 proto static metric 101
...

$ ip rule

0:  from all lookup local
32764:  from IP2 lookup isp2
32765:  from IP1 lookup isp1
32766:  from all lookup main
32767:  from all lookup default

SNAT is applied for the traffic originating from eth1:

iptables -t nat -A POSTROUTING -i eth1 -o eth0 -j SNAT --to-source IP1

Current situation:

- All services forwarded from eth0 are working normally.
- All traffic originating from intranet passes out and back normally.
- All the attempts to access services from eth2 time out.

There are no obvious hints in /var/log/messages (such as complaints 
about "martian IPs").

I am somewhat at a loss here, all the pieces of advice would be very 
welcome.

-- 
Sincerely,

Konstantin Boyandin