Very Strange Remote Buffering Problem

[PR-0927]

Hi all, I've been having a rough time getting my Jellyfin server (TrueNAS Scale) up and running as a proper replacement for my Plex server (UnRAID) I've been running for a while. I would very badly like to transition to Jellyfin and ditch Plex for myriad reasons, but one very strange issue has become a thorn in my side - something which does NOT affect my Plex setup at all.

Hoping for assistance. Crossing my fingers for a resolution!

On local LAN, it works PERFECTLY - direct stream and transcoding both.

The thorn - ONE problem that I cannot resolve - when remote streaming it buffers every few seconds, regardless of resolution or codec. Some items go smoothly, some don't. Does not matter if direct stream or transcode.

This was tested by my sister, remotely located - using Roku (TCL TV built-in), Chrome browser on her desktop, and Jellyfin app on her Pixel 7 Pro - same experience on all three. Many settings changed and tested - auto, 4K 120Mbps, 1080p 10Mbps, etc. Transcoding enabled, disabled, forced. Transcoding throttling enabled or disabled.

Always the same problem.

The weirdest wrinkle - if I stream to my phone (Pixel 7 Pro, GrapheneOS) or my wife's phone (Pixel 7 Pro, Google's Android) - on data (WLAN works perfectly, of course), but I toggle on ProtonVPN (on our phones) - it suddenly functions PERFECTLY. Without that, buffering happens. This is regardless of resolution or codec.

Meaning even 4K 120Mbps, it works perfectly in that scenario.

This is the opposite of what I expected - an additional tunnel layer surely should SLOW things down, right?

Here's the relevant hardware and software background:

Hardware

• TrueNAS Scale Server (New)
  • AMD EPYC 9124
    
  • 128GB DDR5 ECC RAM
    
  • 2x240GB SATA SSDs - Boot Drives – Mirrored
    
  • 2x1TB NVMe (PCIe 4.0) M.2 SSDs - Apps/VM Drives – Mirrored
    
  • 4x8TB NVMe (PCIe 3.0) U.2 SSDs - Media Storage - RAIDZ1 (24TB Usable)
    
  • Nvidia Quadro P400 - Transcoding GPU
    
  • 10Gb PCIe NIC - Intel X520 Chipset - SFP+
    

• UnRAID Server (Old)
  • AMD Ryzen 3700X
    
  • 64GB DDR4 RAM
    
  • 1TB NVMe (PCIe 4.0) M.2 SSD – Cache
    
  • 4x8TB SATA HDDs - Media Storage - Double Parity (16TB Usable)
    
  • No GPU (Previously Nvidia GTX 1080)
    
  • 10Gb PCIe NIC - Intel X520 Chipset - SFP+
    

Software

• TrueNas Scale Server (New)
  • Jellyfin - TrueCharts Repository - ingressed Through Traefik
    
  • Traefik - TrueCharts Repository
    
  
  
• UnRAID Server (Old)
  • Plex - Community Apps - plexinc Repository - NOT Through Reverse Proxy, Exposed Though Port 32400
    
  • Nginx Proxy Manager - Community Apps - jlesage Repository - Currently Disabled - Not Used for Plex
    

Other

• Domain's DNS Through Cloudflare - DNS Only (NOT Proxied) - Cache Bypass
  
• Custom OPNSense Firewall/Router - Unbound DNS (DNSSEC and Blocklists Currently Disabled) - Using 1.1.1.1. and 1.0.0.1 - "Host Override" for My Domain (media.mydomain.tld) and Looal Server IP (192.168.0.50)
  
• ISP - Spectrum - 940Mbps Download, 35Mbps Upload (40Mbps Normally Witnessed)
  

I have experimented with a LOT of different settings. I have sought out help from the support staff for TrueCharts version of Jellyfin. I previously had Quad9 set for my DNS, but thought maybe it was the culprit. Guess not.

Furthermore, I have also used the ix version of Jellyfin on that TrueNAS Scale server and had the same experience.

Happy and willing to provide more information or screenshots as needed. The logs don't seem to be very helpful - only random mentions of "slow HTTP response" - but this happens both when it doesn't work right remotely, and when it does (ProtonVPN toggled on phones).

Super grateful in advance for help on this!

[Voltage6905]

I have a similar issue. I am running Jellyfin in docker on my Intel based OMV machine. Local direct play/transcoding is flawless but I have stuttering via remote transcoding regardless of quality. I am using a wireguard VPN to stream remotely but I have also used NGINX proxy manager with the same problem.

For me the issue is not consistent though. It usually only buffers around 9 PM - 12 AM where I am. I suspect what's happening is the traffic is being throttled somewhere between my house and my parents house. With Plex I believe traffic bounces through their servers somewhere and I don't think that traffic is throttled. Just a theory though since I've been troubleshooting this issue for a year now with no answers.

[PR-0927]

Oh interesting - yeah even the level of buffering has not been consistent in my tests. I tried routing my Jellyfin server through "Gluetun" (VPN option in TrueCharts apps on TrueNAS Scale) - via my ProtonVPN credentials (OpenVPN protocol, not the WireGuard one) - no difference.

Tried connecting via WireGuard to my home network - also no difference.

So I think your theory might be right. Which is super unfortunate, because I don't want to stick to Plex...

[cloud9_development]

Yeah, I would assume the issue is traffic for plex is generally routed through their jump servers, whereas Jellyfin is 100% direct connect. Not to say there aren't other factors at play that could cause you issues, but off the top of my head that's the main difference between the two services

---

One thing you can try to do is run various network testing over the course of a few days to see if there's any packet loss, speed fluctuations, throttling, etc between the two endpoints (your server and whereever you're experiencing issues from).

In my case, 90% of the time I have issues, its due to packet loss (my ISP hasn't been the best and I'm stuck with them..)

[Voltage6905]

One thing you can try to do is run various network testing over the course of a few days to see if there's any packet loss, speed fluctuations, throttling, etc between the two endpoints (your server and whereever you're experiencing issues from).

In my case, 90% of the time I have issues, its due to packet loss (my ISP hasn't been the best and I'm stuck with them..)

How did you go about diagnosing that? Whenever I encounter issues I monitor WAN bandwidth on the server side and there is always plenty of bandwidth to spare. Regardless though, any ideas on how to get around this if throttling is the issue? I've run out of ideas with this.

[cloud9_development]

One thing you can try to do is run various network testing over the course of a few days to see if there's any packet loss, speed fluctuations, throttling, etc between the two endpoints (your server and whereever you're experiencing issues from).

In my case, 90% of the time I have issues, its due to packet loss (my ISP hasn't been the best and I'm stuck with them..)

How did you go about diagnosing that? Whenever I encounter issues I monitor WAN bandwidth on the server side and there is always plenty of bandwidth to spare. Regardless though, any ideas on how to get around this if throttling is the issue? I've run out of ideas with this.

For Internet issues (packet loss for example) I wrote a simple script that runs a ping test of 500 packets every 7 minutes (using crontab) and saves to a log file. You could do something similar. 

There's also some online tools you can use (not sure off the top of my head though) that can look at ingress traffic and give you packet statistics hitting your network

[PR-0927]

For those curious, I kept suspecting something with my router. Turns out it was the culprit, almost certainly. You can read what all I tried and the eventual resolution:

https://forum.opnsense.org/index.php?topic=35487.0

[bitmap]

Server is borked. Care to share a brief summary?