Can't ping or access unraid unless I remove entry from arp table

[bmrowe]

I have a super weird issue where my unraid server will frequently become inaccessible from any computer on the network. The web gui will not be accessible and pings will fail. All dockers are still accessible, only unraid itself it impacted. This will only be resolved by either removing the entry in the arp table for the ip address of the unraid server, or turning wifi off and back on for a client. I have also verified that the router cannot ping unraid during these outages.

 

I've made a few changes lately - adding nginx as an internal proxy to avoid memorizing ip/port combos, and sonarr/radarr/prowlarr/overseerr. I don't see how either would be impacting this.

 

Thoughts?

Attaching diagnostics.

unraid-diagnostics-20231107-1714.zip

Edited November 9, 2023 by bmrowe

[dlandon]

Start by setting up a DNS server for your Unraid server.  This is normally your router.  You don't have a DNS server defined.

[bmrowe]

2 hours ago, dlandon said:

Start by setting up a DNS server for your Unraid server.  This is normally your router.  You don't have a DNS server defined.

Thanks for chiming in. I might be missing a setting, but I have this configured in network:

 

I'll do a basic rundown of my network too:

192.168.1.1 is my router
192.168.1.3 is an adguard home docker running as DNS
192.168.1.24 is a pi hole docker running as secondary DNS
192.168.1.13 is an unbound docker running as upstream DNS for both adguard and the pi hole.

Here is the config in the router:

 

Finally, I will also add that when unraid is unreachable on one pc, it may still be reachable on another. 

Edited November 10, 2023 by bmrowe

[bmrowe]

An update - I changed several things:

-I updated my DHCP server to only share out IPs in the .200-250 range thinking that perhaps there was an IP address conflict despite having IPs reserved for most clients including unraid.

-I have had issues with dockers talking to eachother and had added a script to check if the unraid shim was in place, and if not, add it. I disabled this script.

-I disabled netdata, cadvisor, and the prometheus exporter for adguard. I did this primarily because there was so much network chatter it was hard to troubleshoot w/ wireshark

 

None of these changes solved the problem. However, I noticed CPU use was pretty high and the top two processes were firefox-bin being launched from the /tmp/ directory. I had no clue what this process was, so I decided to restart.

Since restarting, this issue has not reoccured once. I'm not sure which change, if any, fixed this problem - but I wanted to share the latest.