Centos – /etc/sysconfig/network is ignored on CENTOS7

centosnetworking

I have a server with two NIC's. The default gateway has to be on a specific one, therefore I configured the GW via /etc/sysconfig/network. But obviousely, this is completely ignored by the NetworkManager

[root@node1 ~]# cat /etc/sysconfig/network
GATEWAY="10.100.20.1"
[root@node1 ~]# systemctl restart network
[root@node1 ~]# ip r s
default via 10.100.7.1 dev ens160  proto static  metric 1024
10.100.7.0/24 dev ens160  proto kernel  scope link  src 10.100.7.31
10.100.20.0/22 dev ens192  proto kernel  scope link  src 10.100.20.204

Any hints?

EDIT: Here are the interface configurations:

[root@node1 ~]# cat /etc/sysconfig/network-scripts/ifcfg-ens192
DEVICE="ens192"
ONBOOT=yes
NETBOOT=yes
UUID="e634ce78-f5ea-4b3c-b3ea-1e40ef0b11ac"
BOOTPROTO=none
IPADDR="10.100.20.204"
NETMASK="255.255.252.0"
GATEWAY="10.100.20.1"
HWADDR="XXX"
TYPE=Ethernet
NAME="ens192"
DNS1="193.12.103.2"

[root@node1 ~]# cat /etc/sysconfig/network-scripts/ifcfg-ens160
DEVICE="ens160"
ONBOOT=yes
NETBOOT=yes
UUID="bee8e636-80b4-45f3-9b74-eb9eb0fed38d"
BOOTPROTO=none
IPADDR="10.100.7.31"
NETMASK="255.255.255.0"
GATEWAY="10.100.7.1"
HWADDR="XXX"
TYPE=Ethernet
NAME="ens160"
DNS1="193.12.103.2"

And here's a part from messages that probably causes the Issue:

[root@node1 ~]# cat /var/log/messages | grep NetworkManager | grep "Nov  9" | grep "as default for IPv4" -A5 -B5
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> (ens192): device state change: ip-config -> ip-check (reason 'none') [70 80 0]
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> Activation (ens192) Stage 5 of 5 (IPv4 Commit) complete.
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> (ens160): device state change: ip-check -> secondaries (reason 'none') [80 90 0]
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> (ens192): device state change: ip-check -> secondaries (reason 'none') [80 90 0]
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> (ens160): device state change: secondaries -> activated (reason 'none') [90 100 0]
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> Policy set 'ens160' (ens160) as default for IPv4 routing and DNS.
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> Activation (ens160) successful, device activated.
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> (ens192): device state change: secondaries -> activated (reason 'none') [90 100 0]
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> NetworkManager state is now CONNECTED_GLOBAL
Nov  9 12:24:36 logging-02 NetworkManager[631]: <info> Activation (ens192) successful, device activated.
Nov  9 12:24:41 logging-02 NetworkManager[631]: <info> startup complete

Any idea what kind of "Policy" is mentioned there?

Best Answer

This worked:

[root@node1 ~]# cat /usr/share/doc/initscripts-9.49.17/sysconfig.txt | grep DEFROUTE
DEFROUTE=yes|no (set this interface as default route? yes is default)

I still believe that ignoring /etc/sysconfig/network is a bug in CentOS Linux release 7.0.1406 (Core)

Related Solutions

Centos 6.2 Fresh ‘Basic Server’ install networking issues

Your routes and configuration look fine.

$: route -n
Destination // Gateway // Genmask // Flags // Metric // Ref // Use // Iface
66.*.*.0       0.0.0.0    255.255.255.248   U   1003      0   0     0  em2
0.0.0.0        66.*.*.1   0.0.0.0           UG     0      0   0     0  em2

The first route, 66.*.*.0/29 with gateway 0.0.0.0 tells your computer to use interface em2 and then make an arp request to find the hardware address of the host you're trying to reach. This is a "connected" route.

The second one is the default route, pointing at your default gateway through em2. If you need to send a packet in another network than 66.*.*.0/29, your computer will make an arp request to find 66.*.*.1 and then send the packets to it.

The only thing in your configuration that could be an issue is the NM_CONTROLLED=yes statement in /etc/sysconfig/network-scripts/ifcfg-em2. This tells the system that this interface is controlled by NetworkManager. This could interfere with your static configuration.

However, even without any default gateway you should be able to ping and ssh from the 66.*.*.0/29 subnet to your machine.

Check layer 1 first, and ensure that the cable is plugged on each side. Use leds on nic and switch, and check if the system sees it correctly:

# mii-tool
eth0: negotiated 1000baseT-FD flow-control, link ok

Then verify if any iptables are dropping the packets. Use iptables -L or iptables-save to check for any rules, and iptables -D <rule> to delete them. Pay attention to the default policy.

Also, on some systems, NetworkManager can configure ufw automatically, and I've had issues with static interface configuration that wasn't seen by NM and hence blocked by ufw.

Redhat – Centos multiple NICs routing issue

For any network device which is not my default gateway, I usually set the default route flag to no:

DEFROUTE="no"

This seems to work without any issues for my servers with multiple network interfaces. If you then restart your network service or interfaces, you should be able to check the routes to see that this is actually working:

/sbin/route -n

Hopefully this helps.

Best Answer

Related Solutions

Centos 6.2 Fresh ‘Basic Server’ install networking issues

Redhat – Centos multiple NICs routing issue

Related Topic