We have all been there: you click “Update” on a stable Unraid build, and upon reboot, your most critical service, Plex, refuses to start.
After updating to Unraid 7.2.3, my lscr.io/linuxserver/plex container went dark. The logs were unavailable, and the symptoms pointed toward a deeper configuration issue rather than a container failure. Here is the breakdown of how a DNS loop and an NVIDIA driver mismatch created a perfect storm, and how to fix it.
The Environment
- OS: Unraid 7.2.3
- Container:
lscr.io/linuxserver/plex - GPU: NVIDIA GTX 1050 Ti
- Key Plugins: NVIDIA Driver, Community Applications
Symptoms
The first sign of trouble was a total breakdown of the Unraid ecosystem connectivity:
- Plex container stuck in a “Stopped” state.
- Community Applications (Apps store) failed to load.
- NVIDIA Driver plugin showed limited version options.
- System notifications warned:
Warning: Cannot reach GitHub. Note that this may be a temporary issue with GitHub or your internet connection.
Step 1: The Connectivity Trap
My first instinct was to check the internet. A simple ping to a public IP proved the pipe was open:
ping -c 4 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
64 bytes from 8.8.8.8: icmp_seq=1 ttl=116 time=16.6 ms
64 bytes from 8.8.8.8: icmp_seq=2 ttl=116 time=14.5 ms
64 bytes from 8.8.8.8: icmp_seq=3 ttl=116 time=27.2 ms
64 bytes from 8.8.8.8: icmp_seq=4 ttl=116 time=10.5 ms
--- 8.8.8.8 ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3003ms
However, the “Cannot reach GitHub” error suggested a name resolution failure. I tested DNS specifically:
ping -c 4 github.com
ping: github.com: Name or service not known
Step 2: The AdGuard DNS Loop
The root cause was my DNS architecture. My Unraid Host OS was pointing to my AdGuard Home container (192.168.1.199) for DNS.
The Gotcha: Containers like AdGuard are not available early in the Unraid boot sequence. Because the host could not resolve GitHub, it could not verify plugins or pull the necessary dependencies for the Plex container to initialize.
The Solution: I updated the Unraid Network Settings to use public DNS servers (1.1.1.1 and 8.8.8.8) for the Host OS. This ensures the server can always “phone home” during boot, while AdGuard continues to handle DNS for the rest of the LAN.
Once applied, name resolution was restored:
ping -c 4 github.com
PING github.com (140.82.114.3) 56(84) bytes of data.
64 bytes from lb-140-82-114-3-iad.github.com (140.82.114.3): icmp_seq=1 ttl=241 time=50.9 ms
64 bytes from lb-140-82-114-3-iad.github.com (140.82.114.3): icmp_seq=2 ttl=241 time=47.7 ms
64 bytes from lb-140-82-114-3-iad.github.com (140.82.114.3): icmp_seq=3 ttl=241 time=43.4 ms
64 bytes from lb-140-82-114-3-iad.github.com (140.82.114.3): icmp_seq=4 ttl=241 time=36.2 ms
--- github.com ping statistics ---
4 packets transmitted, 4 received, 0% packet loss, time 3003ms
Step 3: NVIDIA Driver Mismatch
Fixing the DNS was a prerequisite for fixing the GPU. Before the DNS fix, the NVIDIA Driver plugin could not reach the internet to “see” the full list of available drivers. It was only showing v590.48.01.
Once the host had internet access, the plugin refreshed and displayed the legacy options. My GTX 1050 Ti is a legacy card in the eyes of modern drivers. It requires the 580 legacy branch to maintain stable hardware acceleration. I updated the driver to v580.126.09 via the plugin settings.
Step 4: Verification
I ran the NVIDIA System Management Interface in the Unraid terminal to confirm the OS recognized the card:
nvidia-smi
Result: GTX 1050 Ti recognized, driver v580.126.09 loaded.
Plex started successfully, and GPU transcoding was once again visible inside the container.
Summary of Root Causes
- DNS Circular Dependency: The host relied on a container that had not started yet to resolve addresses.
- Plugin Failure: Without DNS, the NVIDIA plugin could not fetch the legacy driver list.
- Driver Mismatch: The OS update defaulted to a driver version incompatible with legacy GPU hardware.
Lessons Learned
- Host DNS Isolation: Never point your Host OS DNS solely to a containerized service (AdGuard or Pi-hole) running on that same host.
- Ping IP vs. Ping Domain: Always verify DNS resolution separately from raw internet connectivity.
- Legacy Hardware Vigilance: When running older GPUs, always double check your driver versions after a kernel or OS update.
By decoupling the host DNS and pinning the correct legacy driver, Plex was back up and transcoding in minutes.
