Changed default NIC, works, but when dockers start, networking crashes

[bobo89]

I have an intel onboard NIC, and a 4 port Realtek addin card. When I first setup the server the default port was on the realtek card.

 

Today I decided to change the default port to be the Intel onboard card. I stubbed the whole addin card, and nuked the networking file. A few reboots later, the GUI comes up, and the exercise seems to have worked.

 

The instant I enabled Docker containers (edit, it's VM manager) networking goes down. I now reboot, the webGUI comes up, and the second vm manager starts  loading networking goes down (can't ssh, or view webgui)

 

The last few lines of syslog can be seen attached. Any pointers? 

Edited October 28, 2019 by bobo89
clariyfying info. vm manager starts and network crashes, not docker

[Vr2Io]

Should be wrong nerwork setting.

To clear ( reset ) network config, you may delete the network.cfg in flash folder /config/ and reboot again.

[bobo89]

Should be wrong nerwork setting. To clear ( reset ) network config, you may delete the network.cfg in flash folder /config/ and reboot again.

 

I deleted the file and rebooted. The server comes up, can access the webGUI, my VM autoboots, but when it does the network gets killed again. I managed to beat the race condition and disable VMs which preserves the network state and allows me to interact with the webGUI. Any other thoughts? 

Screenshot may be relevant..it looks like the VM manager is trying to connect to virtbridge which I explicitly told to just use br0, not the virtual bridge.

 

 

[Vr2Io]

Any dirty config in docker / VM about Realtek NIC ?

 

You stub the NIC for what service.

 

Edited October 21, 2019 by Benson

[bobo89]

Any dirty config about stub Realtek NIC in docker / VM ?

 

Looking into docker.log is empty. When I enabled VMs I didn't see any mention of the Realtek card in /var/log/libvirt logs. where else could I look ? 

 

Looks like the config file is still trying to use virbr0, which isn't what I have set in the gui..

Edited October 21, 2019 by bobo89

[Vr2Io]

Suggest try disable stub NIC and check problem fix or not first.

 

May be issue not on config or stub, you also need check how IP allocate ( NIC change with dynamic IP ), it may cause "network down" too.

 

Edited October 21, 2019 by Benson

[bobo89]

48 minutes ago, Benson said:

Suggest try disable stub NIC and check problem fix or not first.

 

May be issue not on config or stub, you also need check how IP allocate ( NIC change with dynamic IP ), it may cause "network down" too.

 

So I disabled the stub, and this brings back exactly why I'm where I am. With the realtek card plugged in I can't get the onboard NIC to get an IP no matter what I try. 

 

After disabling the stub, I boot with no IP, no network.cfg (I deleted it to regenerate, but it doesn't). Syslog is full of references of trying to address eth1,2,3, which is the realtek card, but not taking eth0.

 

Edit: i've managed to get eth0 as the primary interface, and working without stubbing. However, the instant a VM boots, it kills all networking on the host.

Edited October 21, 2019 by bobo89

[Vr2Io]

4 hours ago, bobo89 said:

Edit: i've managed to get eth0 as the primary interface, and working without stubbing. However, the instant a VM boots, it kills all networking on the host.

So, in simple say, no matter stub Realtek NIC or not, once docker/VM start, all network down.

 

How about if no cable connect to Realtek? If problem persists, then I will try unplug Realtek from system to verify does all resume normal.

 

Does onboard NIC with Realtek NIC have work previously or it just a new setup.

 

You may provide diagnostic and detail network setting, docker/VM setting to get others advice.

 

Edited October 22, 2019 by Benson

[bobo89]

30 minutes ago, Benson said:

So, in simple say, no matter stub Realtek NIC or not, once docker/VM start, all network down.

 

How about if no cable connect to Realtek? If problem persists, then I will try unplug Realtek from system to verify does all resume normal.

 

Does onboard NIC with Realtek NIC have work previously or it just a new setup.

 

You may provide diagnostic and detail network setting, docker/VM setting to get others advice.

 

I have tried with cable in realtek, and without. I have managed to get onboard NIC working with random cable plugged into the first port of the realtek nic.

 

This is my first go with unraid. Initially when i setup the server I wasn't able to get the onboard NIC to work with the realtek card, so I just used the first port of the Realtek card, as that's the only combo UNraid would work with. However for my use case I need those 4 ports passed through for something else.

 

I tried to unplug the Realtek card. I boot normally and receive an IP on the Onboard nic. The instant I enable VM manager, and a VM starts booting, host network goes down. Attached is a diagnostic of before I enable VM manager and without the realtek card plugged in.

tower-diagnostics-20191022-0151.zip

Edited October 22, 2019 by bobo89

[Vr2Io]

You have config "Preserve user defined networks:", may be you could try set to "no" ( docker setting advance view )

 

 

Edited October 23, 2019 by Benson

[bobo89]

9 hours ago, Benson said:

You have config "Preserve user defined networks:", may be you could try set to "no" ( docker setting advance view )

 

 

Tried that. After starting VM Manager, network goes down. Attached is diagnostics after the crash.

tower-diagnostics-20191023-1824.zip

[bobo89]

On 10/23/2019 at 6:27 PM, bobo89 said:

Tried that. After starting VM Manager, network goes down. Attached is diagnostics after the crash.

tower-diagnostics-20191023-1824.zip 113.25 kB · 4 downloads

Look for feedback on this @limetech @RobJ

[bobo89]

The solution here was to disable autostart on all VMs and just open the config and save it again. This seems to have resolved a stale config that was present.

Sent from my SM-N960W using Tapatalk