[6.12.4] containers with br0/eth0 network fail to start after reboot/docker service restart

Below is my docker setting and compose file:

I use macvlan to assign dedicated IP and mac .For network control purpose.

br0 as main network and traefik for external access.

Edited September 15, 2023 by nukecat

I change my networtk and docker setting as relase note suggested.And it still failed to auto start even after docker service restart!

here is my diagnostics.

 endurance-diagnostics-20230919-1351.zip

This looks more like a configuration error.

I dont know how your compose is set up, but given that you dont have preserve custom networks on, i assume traefik is creating the bridge traefik.

Since Aria2 and qbittorrent start before traefik, this is to be expected. Change the autostart order.

16 hours ago, Mainfrezzer said:

This looks more like a configuration error.

I dont know how your compose is set up, but given that you dont have preserve custom networks on, i assume traefik is creating the bridge traefik.

Since Aria2 and qbittorrent start before traefik, this is to be expected. Change the autostart order.

I think it's not the autostart order.

Every time unraid reboot or disable and enable docker service,br0 or eth0  changed its id,therefore container aria2 cannot found the correct network by network id.

Docker network traefik didnt created by container traefik. Its a external network which I manually created it by [docker network create traefik] months back and it just works fine before 6.12.4 upgrade.

 

I did a experiment last nignt and confirmed the changing of docker network id.Unraid did recreate network eth0(or br0 based on config)  when starting docker service.

Ive spend some time on it and youre indeed correct, this is an issue. Sadly not unraid specific, it a compose issue, it just wasnt noticed before. 

"docker-compose up --force-recreate"

20 hours ago, Mainfrezzer said:

Ive spend some time on it and youre indeed correct, this is an issue. Sadly not unraid specific, it a compose issue, it just wasnt noticed before. 

"docker-compose up --force-recreate"

That's odd, everything works fine before 6.12.4 upgrade and docker compose is just a way to control and deploy containers.

Container created by 'docker run' command with br0 or the0 network should facing the same problem (ie containers failed to start because network id changing).

I think the recent change to fix for macvlan call traces changed how unraid create and control docker network,causing it's id changing.

Hope it will be fixed soon.

i placed a pull request for the plugin which will fix the issue.

In the meantime you can fix the compose issue by downloading the compose.manager-package-2023.09.13.txz from your usb. Get the "compose.sh" from 

compose.manager-package-2023.09.13.txz\compose.manager-package-2023.09.13.tar\usr\local\emhttp\plugins\compose.manager\scripts\

modify the line 57 from
 

eval docker compose $files -p "$name" up -d 2>&1

to
 

eval docker compose $files -p "$name" up --force-recreate -d 2>&1

put the compose.sh back into the archive. Let, hopefully youre using 7zip, 7zip rebuild the archive.

Delete the original .txz from the usb, place the modified on the usb (hopefully with dynamix file manager) and then open the terminal 

 

md5sum /boot/config/plugins/compose.manager/compose.manager-package-2023.09.13.txz

copy the md5 sum that appears and modify the md5 checksum on line 8 in the compose.manager.plg within the /config/plugins folder of the usb.

It should be <!ENTITY packageMD5 "771a0a5b9d3f8b002ccaca86a38af635"> if i did manage to not sneak in some spaces.

And then reboot.

Edit: If youre willing to trust me, here are the 2 files so you dont have to do it manually.plugins.zip

Edited September 23, 2023 by Mainfrezzer