Jellyfin performance on Synology

[Sabarok]

Jellyfin is running impossibly slow on my NAS. It takes minutes to open the home page or to go through the dashboard. I've even had it take an hour to open while the scan was happening. I am wondering if there might be some settings I can change that might improve performance. The resources app on the synology is showing everything is waiting on IO. Memory usage is 50%, and the CPU is idling.

The resources I've found, including on this forum, are all about running synology inside a container, but there's a jellyfin community package now available that I'm using, so it's running directly on Synology.

I've just tried installing jellyfin onto my pi4 using the NAS as a network drive, and performance there is plenty fast.

I'm using the Synology DS220j. It's not a strong machine, but it's not doing anything else. Videostation was retired in the latest version, so that's been uninstalled and I've been looking for a replacement.

To save people searching, these are the specs of the 220j:
CPU: Realtek RTD1296 4-core 1.4 GHz
Memory:  512 MB DDR4
Drives: 2x SATA running in mirror RAID

It is running 10.10.1-19

Are there some settings I can try, or is the DS220j simply too underpowered for jellyfin? I'm quite surprised at the vast performance difference between the synology and pi 4, and I'm not talking about playing videos. Just opening the page and looking at the dashboard to configure it.

I could understand there being some speed difference, but the pi4 opens in <1s while Synology is taking minutes.

[gnattu]

Everything is waiting on IO means your HDD is just too slow for database tasks and Jellyfin is quite IO heavy when doing library scanning.

On the CPU side, RTD1296 is Raspberry Pi 3 level of performance, which is much slower than the Pi 4.

[Sabarok]

The library scan when done locally uses so much IO that the NAS is unusable, so I've set it to only scan after midnight. It taking minutes to open is when it's not doing a library scan and all scheduled tasks are dormant.

[kandykarter]

I used to run Jellyfin on my old DS220+ and as my library grew, it became unusable, between the underpowered CPU and the Jellyfin DB running on a hard drive instead of an SSD. I'd highly recommend getting a little N100 mini pc to run jellyfin and just using the NAS for storage.

[Sabarok]

I used to run Jellyfin on my old DS220+ and as my library grew, it became unusable, between the underpowered CPU and the Jellyfin DB running on a hard drive instead of an SSD. I'd highly recommend getting a little N100 mini pc to run jellyfin and just using the NAS for storage.

I have considered getting a N100 micro PC, though I did see a Ryzen 7 5700U at a tempting price. If I got one it would be one with drive bays so I could use it to replace both my synology and pi in the living room.

One thing I have noticed is that my IO is entirely being used by jellyfin. I turned jellyfin off, and utilization went from 100% to 2%. Jellyfin is telling my all scheduled tasks were dormant, so I'm wondering if there might be a bug with how jellyfin works on synology. I had tried a container last week, but it had the same issues. I did notice that trying to "shutdown" jellyfin from the control panel didn't work and I had to do it from Synology.

I think my next step will be to try and remove the libraries from jellyfin to see if I can get io under control, and add the libraries more piecemeal.

[TheDreadPirate]

If it was busy generating trick plays or chapter image extraction, those usually may appear in the dashboard as an active job. But your log would show what activity is going on. Regardless, the disk IO for a database easily saturates a HDD.

What you can do in the meantime is to limit the number of library scan threads.

Dashboard > General. Set both "Parallel library scan tasks limit" and "Parallel image encoding limit" to 1.

If you intend to keep the NAS, I'd choose a mini PC with an N100 over a 5700U. The media engine in the N100 is MUCH better than the 5700U.

If you intend to build a proper server, including having storage be locally attached, we'd need a budget and intended usage to give more specific recommendations.

[Sabarok]

If it was busy generating trick plays or chapter image extraction, those usually may appear in the dashboard as an active job.  But your log would show what activity is going on.  Regardless, the disk IO for a database easily saturates a HDD.

What you can do in the meantime is to limit the number of library scan threads.

Dashboard > General.  Set both "Parallel library scan tasks limit" and "Parallel image encoding limit" to 1.

If you intend to keep the NAS, I'd choose a mini PC with an N100 over a 5700U.  The media engine in the N100 is MUCH better than the 5700U.

If you intend to build a proper server, including having storage be locally attached, we'd need a budget and intended usage to give more specific recommendations.

I came here to report exactly this.

I removed all the libraries, and IO was still high. I had tried changing the parallel setting right there, but it wouldn't let me. I restarted jellyfin, and IO was still high and I couldn't access jellyfin at all. I uninstalled deleting all app data to start fresh. The first thing I did was to set parallel to 1. Adding the first library and IO is high, but jellyfin remains responsive and the scan is happening quickly.

What I think happened is the same thing you are pointing out. This synology is a 4-core, so jellyfin might have been trying 3 or 4 parallel scans. It also wasn't able to control them so even when it tried to stop, the scan sub-processes were still running in the background. Restarting jellyfin was also restarting those scans. Even when I had tried to change the configuration to run nightly, it was still jammed from previous attempts.

Also, thanks for the tip on the N100. I had heard the N100 had a great media encoder, but it's good to know the comparison to the 5700U.

[thombean]

I ran into the same issues on my Synology DS923+, with Jellyfin running, about 80% of my RAM is full and the NAS becomes unresponsive. 4 GB RAM is just not enought for Jellyfin. So either I upgrade my RAM or stick to using Navidrome and Infuse via SMB. Plex is also lighter on ressources.

[Sabarok]

To give a final update: I have decided this is a failed endeavor. Even though the dashboard was super fast, any time jellyfin needed to access the database the IO would spike and it would be completely unresponsive. I have removed jellyfin from the NAS. It now sits on the pi4. I'm currently trying to figure out how to upgrade it from 10.8 to 10.10, but that would be for a different thread.

[eetaylog]

Does your Nas have the ability to add M.2 SSD cards? You can set them up as a cache.

You could also look into upgrading your RAM.