Linux policy routing is supported via a policy rule database. In the database there are several tables which have their own priorities(TABLE_ID is an eight bit integer).
the less the TABLE_ID is, the higher its priority is.

you are able to see the full rules list via the following command:
$ip rule list
0:	from all lookup local
32766:	from all lookup main
32767:	from all lookup default

you can also specify table's alias to make it looks more meanningful by editing the /etc/iproute2/rt_tables. 

when kernel is making route decisions, it first looks up the rule database basing on its packet characteristics to match a specific table, and then do routing
inside that table.
****. Policy strutcture. 
    There are couple of criterias for you to match your traffic patterns. for example, you want to direct the traffic flows with source port 8080 to 
    an asymmetric gateway 130.140.150.100 as next hop, and this violates the routes in you main table. so you can create a rule table to do customized routing yourself:
        $sudo ip rule add sport 8080 table 100
        $sudo ip route add default via 130.140.150.100 table 100 
    
    besides, there are other selectors for you to create rules, below is the syntax:
    SELECTOR := [ not ] [ from PREFIX ] [ to PREFIX ] [ tos TOS ] [ fwmark FWMARK[/MASK] ]
            [ iif STRING ] [ oif STRING ] [ pref NUMBER ] [ l3mdev ]
            [ uidrange NUMBER-NUMBER ]
            [ ipproto PROTOCOL ]
            [ sport [ NUMBER | NUMBER-NUMBER ]
            [ dport [ NUMBER | NUMBER-NUMBER ] ]
****.A case of asymmetric routing.
     the topology diagram is given as below:
               +---------------------------------------------------------+
               | Host(default namespace)                                 |
               |                                                         |
               |        +-----------------------------------+            |
               |        |ns_demo       (web service:8080)   |            |
               |        |      10.0.0.2/24                  |            |
               |        |        +            172.13.14.2/24|            |
               |        |        |                 ^        |            |
               |        |        |                 |        |            |
               |        |        |                 |        |            |
               |        +-----------------------------------+            |
               |                 |                 |                     |
               |                 |                 |                     |
               |            Downstream Link        | Upstream Link       |
               |                 |                 |                     |
               |                 v                 |                     |
               |           10.0.0.1/24             |                     |
               |                                   +                     |
               |                                   172.13.14.1/24        |
               |                                                         |
               +---------------------------------------------------------+
     traffic from the host to the namespace flows via upstream link, the reponse follows another path. here are steps to present the demo:
        $sudo ip netns add ns_demo
        $sudo ip link add name if_uplink_host type veth peer if_uplink_demo
        $sudo ip link set if_uplink_host up
        $sudo ip link set if_uplink_demo netns ns_demo
        $sudo ip netns exec ns_demo ip link set if_uplink_demo up
        $sudo ip addr add 172.13.14.1/24 dev if_uplink_host
        $sudo ip netns exec ns_demo ip addr add 172.13.14.2/24 dev if_uplink_demo
    now the upstream link is ready, you are able to test the connectivity between two namespaces. let setup downstream link.
        $sudo ip link add name if_dnlink_host type veth peer if_dnlink_demo
        $sudo ip link set if_dnlink_host up
        $sudo ip link set if_dnlink_demo netns ns_demo
        $sudo ip netns exec ns_demo ip link set if_dnlink_demo up
        $sudo ip addr add 10.0.0.1/24 dev if_dnlink_host
        $sudo ip netns exec ns_demo ip addr add 10.0.0.2/24 dev if_dnlink_demo
    Now downstream link is also ready, the are two routes in ns_demo's main table:
        $sudo ip netns exec ns_demo ip route show table main
        10.0.0.0/24 dev if_dnlink_demo proto kernel scope link src 10.0.0.2
        172.13.14.0/24 dev if_uplink_demo proto kernel scope link src 172.13.14.2
    That's quite expected. next we are going to configure the policy:
        $sudo ip netns exec ns_demo ip rule add sport 8080 table 100
        $sudo ip netns exec ns_demo ip route add default via 10.0.0.1 dev if_dnlink_demo table 100
    To verify by starting a web service on port 8080:
        $sudo ip netns exec ns_demo ip route show table 100
        default via 10.0.0.1 dev if_dnlink_demo
        $sudo ip netns exec ns_demo python3 -m http.server 8080

    Finally, we are going to capture packets:
        $sudo tcpdump -i if_uplink_host -n
        tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
        listening on if_uplink_host, link-type EN10MB (Ethernet), capture size 262144 bytes
        11:48:44.343015 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [S], seq 138105168, win 64240, options [mss 1460,sackOK,TS val 2458541685 ecr 0,nop,wscale 7], length 0
        11:48:44.343059 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [.], ack 2845715807, win 502, options [nop,nop,TS val 2458541685 ecr 2638739449], length 0
        11:48:44.343251 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [P.], seq 0:143, ack 1, win 502, options [nop,nop,TS val 2458541685 ecr 2638739449], length 143: HTTP: GET / HTTP/1.1
        11:48:44.344793 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [.], ack 156, win 501, options [nop,nop,TS val 2458541686 ecr 2638739450], length 0
        11:48:44.344855 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [.], ack 1645, win 494, options [nop,nop,TS val 2458541686 ecr 2638739450], length 0
        11:48:44.346043 IP 172.13.14.1.58964 > 172.13.14.2.8080: Flags [F.], seq 143, ack 1646, win 501, options [nop,nop,TS val 2458541688 ecr 2638739450], length 0

        $sudo tcpdump -i if_dnlink_host -n
        tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
        listening on if_dnlink_host, link-type EN10MB (Ethernet), capture size 262144 bytes
        11:48:44.343043 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [S.], seq 2845715806, ack 138105169, win 65160, options [mss 1460,sackOK,TS val 2638739449 ecr 2458541685,nop,wscale 7], length 0
        11:48:44.343266 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [.], ack 144, win 508, options [nop,nop,TS val 2638739449 ecr 2458541685], length 0
        11:48:44.344777 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [P.], seq 1:156, ack 144, win 508, options [nop,nop,TS val 2638739450 ecr 2458541685], length 155: HTTP: HTTP/1.0 200 OK
        11:48:44.344851 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [P.], seq 156:1645, ack 144, win 508, options [nop,nop,TS val 2638739450 ecr 2458541686], length 1489: HTTP
        11:48:44.344892 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [F.], seq 1645, ack 144, win 508, options [nop,nop,TS val 2638739450 ecr 2458541686], length 0
        11:48:44.346067 IP 172.13.14.2.8080 > 172.13.14.1.58964: Flags [.], ack 145, win 508, options [nop,nop,TS val 2638739452 ecr 2458541688], length 0

    SEE!, the TCP stream is splitted unidirectionally on two links.