Docker container adds default route via veth?

[SpiderPigLoki]

Hello everybody,

whenever I start a docker container, it automatically adds a new default route

[alamahant]

Hi please have a look at this
https://docs.docker.com/network/bridge/
in the network stanza try to remove the default gateway line and check if there is a way to add a route instead of the gateway
something like this in /etc/docker/daemon.json

[AlexJGreen]

_

[SpiderPigLoki]

Hello everyone,

thanks for the replies.

1.) I will first check the kernel - judging by this output it can't hurt

[kfiadeg]

Hello there.

I am facing similar issues but more ofter my connection to the Internet is lost when I'm STARTING or STOPPING stack of containers with docker-compose.

It started to happen when I upgraded my kernel from 4.x to 5.x (currently it is 5.4.28).

I have not invented a solution yet. I must always restart my phisical network device completely:

[alamahant]

Maybe you should try with these

[kfiadeg]

Everything on? Hmm, I will give it a try, but this looks like "bruteforce" solution Thanks @alamahant.

I forgot to mention that issue I've described happens on two separate machines and - what is especially driving me mad - it does not happen always. but only from time to time.

I found another thread on this forum which might be related: https://forums.gentoo.org/viewtopic-t-1105886.html

Note: No iptables are configured on the host machine(s).

[SpiderPigLoki]

Hello everybody,

still no success on my end - issue still persists.

I have enabled all of the NETFILTER-options mentioned by alamahant.

[SpiderPigLoki]

Okay played around a bit - without success.

- all iptables and docker stuff was removed (e.g. config files and /var/lib/iptables/rules-save)
- emerge -C iptables app-emulation/docker
- following line in package.use/world :

[AlexJGreen]

_

[SpiderPigLoki]

It appears that I have found something along the lines of a workaround (maybe even solution).

It pains me to write this, but it seems that my beloved OpenRC / netifrc is at fault.

After removing net-misc/networkmanager; emerging openrc without netifrc-use-flag and rebooting solved the issue.

It seems that netifrc added the new device again and again and again, just like coderanger suggested.

For the time being I am going to stick with this solution.


Thanks a lot for all the help.

[AlexJGreen]

_

[kfiadeg]

Nice talk here. It's motivating

So - after what coderanger said - there should be a way also for netifrc subsystem to configure ignored (unmanaged) devices.
So far I have faced only functions preup(), predown(), postup(), postdown(), failup() and faildown() as a possible solution. They might rely on ${IFACE} variable which is set to the interface being brought up/down.

I have not invented the solution yet, though.

I will now dig into https://github.com/gentoo/netifrc/blob/master/doc/net.example.Linux.in

[kfiadeg]

For whom it may concern:

1. I have failed trying to set up netifrc so it would ignore setting default router for certain devices
   
2. I have noticed that also OpenVPN was setting new default route from time to time upon disconnection
   
3. I have used genkernel to generate new kernel - with most (if not all) IP tables features compiled as modules - this helped, overwritting default route stopped to happen
   
4. I have finally created brand new config (make defconfig) for kernel 5.4.48, adjusted it little bit, made sure that all options required by docker are compiled in* and - since then (it is just a few days now) - everything works properly

*) I've done it with help of two supplementary simple community scripts:

• https://gist.github.com/mrueg/cd19a20d2e712f61d2ea
  
• https://termbin.com/lz2k

Both extract informations about kernel requirements from ebuilds of currently installed packages and print them out to the console (the same that show up when installing package). Thank you good people from #gentoo IRC channel on irc.freenode.net !!!

Home it helps.