Really trying to make 6.12.4 work... connectivity/hard resets 😬 is this still a macvlan issue?

[johnny2678]

Background:

• since I started running unraid 3 years ago, I access VMs via Microsoft RDP and access my unraid docker services from that VM over RDP

 

Summary:

• Containers can't be accessed by VMs with VM set to vhost0, but VMs can access services running on other VMs
• VMs can't be access by RDP (MS RDP) with VM set to vbri0, but containers can be accessed by vms
• overnight, CPU load goes to 1000 forcing a hard shutdown

 

I don't like to change my settings once I have things working, but I tried upgrading to 6.12.3 and bumped into the macvlan issues so I rolled back to 6.11.5.  Then, I decided to give 6.12.4 a shot since it was supposed to fix the macvlan issue.

 

Made the changes outlined here:

• bonding = yes
• bridging = no
• host access = yes

 

1st boot on 6.12.4, everything ran fine for a couple of days and I thought the macvlan problem was solved.  Then I woke up this morning to an unresponsive server (see top screenshot).

 

Had to hard reset. 💩

 

When it came back up, I ran into the conditions in the summary above where I couldn't access docker containers from a VM like I have been doing for years.  I really don't want to roll back to 6.11.5 again but might have to.  Hoping someone here can see something that I missed... TIA.

 

Questions:

• What setting do I need to make on 6.12.4 to access docker containers from a VM?
• What is causing my server to go unresponsive forcing a hard reset?

 

 

15620-beast-diagnostics-20230905-0809.zip

Edited September 5, 2023 by johnny2678

[johnny2678]

Just realized this is a Sev1 issue in my house because I have WeeWX running in an unraid VM sending all house temperature data to MQTT running in an unraid docker container.  Been working this way for years but now on 6.12.4 the VM can't see the container. 

 

My AC trips on/off based on that temperature data getting to the MQTT container.  I live in FL. Will have to revert to 11.5

[johnny2678]

I shut down docker/vms services and went through settings again.  Turned off bonding and restarted. I had one nic card unplugged because I need to recable.  Thought bonding would just use the other line, but maybe not. edit: just to clarify, I was running a bond with one cable from 6.8.x and no issues until 6.12.4

 

Either way, VMs can see container services now that I'm just running on ETH0.  crisis averted... for now.  

 

Wondering if this also explains the CPU pegged at 1000?

Edited September 5, 2023 by johnny2678
clarity

[johnny2678]

On 9/5/2023 at 10:43 AM, johnny2678 said:

Wondering if this also explains the CPU pegged at 1000?

 

Nope - CPU pegged again last night and had to hard shutdown.  Back to 6.11.5.