So I’m back with Netflix and Amazon. I’m kind of done with jellyfin at this point. I’ve been recommending it to everyone and honestly I’m not sure that I still would.
Jellyfin as a product simply doesn’t work. There, I said it.
For a media system I’d hope for the basics to be able to play a movie or show with subtitles, selecting audio channels, and scrubbing and it to remember where I was left. I wouldn’t even need posters and all that, I’d get my own subtitles, etc.
I have multiple TVs, but let’s go with the old one with a chrome cast. I use my Android phone to cast a show episode. Scrubbing sometimes works. it works better than before because before it would just break everything to the point of the app requiring a reinstall. At least that’s fixed now but scrubbing still is SLOW, 10 seconds ahead takes 20-30 seconds to do. Worse though, it borks up subtitles by leaving the subtitles that were there when you started scrubbing, and places the new ones on top of that. The only way out is to exit casting completely, kill the app, restart all from scratch. Obviously JF doesn’t continue where I stopped (seriously, requests on that timeframe were made, it can’t be that hard to register that, can it?) and so I need to scrub again which, you know, borks up subtitles. So I’ll start just from the beginning, I can rewatch the same 5 minutes twice but it annoys the hell out of my wife.
Then, subtitles is a mess. The above, but also switching subtitles will cause similar issues. Don’t touch anything when having subtitles!
Then: it’s slow. It’s godunholy slow. I have a 16 core and Rhyzen 5 cpu, 64gb men @3200mhz, 1tb M2 Samsung Evo 989 pro, and ~60tb over 3 exos drives. An AMD rx7800 XT finishes up the config. It’s not the best of the best, but my system, u believe, ranks in the higher ranks of jellyfin installations
Jellyfin cannot play a movie or show without stuttering at least a few times, flat out freezing for minutes during shows, especially in the second episode for that day… If I run transmission in parallel, it just freezes up so much that it’s undoable.
Logs don’t indicate any major issue, I saw a freeze and had all logs on tail and literally saw no messages whatsoever during that freeze. System utilization was near zero.
Wife isnt tolerating jellyfin anymore and now I have Netflix and Amazon accounts again.
I understand it’s open source software, you do what you can, but right now it simply isn’t a system that is for the general public. It can be used by nerds like me who have the patience to deal with all the issues.
Edit: Really? 17 downvotes? I’m trying to tell you that the thing doesn’t work and that I’ve spent plenty of time trying to fix it, reading docs, posting questions, nothing has fixed anything so far but can’t have people talk about that, or what’s wrong?
Probably because your TV needs the media to be transcoded, but not sure what’s wrong there as it’s much smoother here on a Ryzen 1200 (!), media being read from HDD (!), with no help from a GPU. When you skip ahead to a part that hasn’t yet been transcoded, the process needs to start over basically, and that always takes a few seconds, maybe for keeping a minimal buffer but not sure about that, but it really shouldn’t take that much time.
Not sure how you kill the app, but if you actually kill it, it has no chance to save where you were if it wasn’t able to do so continuously. These 2 might be implemented differently by the client, so the “success rate” may differ. Though, so far every time I was casting it (not to chromecast, but to various jellyfin clients) it followed nicely where I was, that may work differently in devices made by certain user-hostile companies. I don’t know that.
That’s just resource exhaustion. If your system is overloaded (even if just the media disk’s queue being full) Jellyfin can do nothing with it. Either move other IO-heavy processes off of that disk, or set them to a lower IO priority (or Jellyfin to higher), or try a different IO scheduler, maybe also apply more of these if needed.
Where do you measure that? Did you look at disk utilization stats? Or how much the CPU has spent on waiting on the disk to respond? This last stat is called “iowait”, you can graph it in e.g. btop/bpytop, or see details in iotop-c/iotop (former of both is better when available, but linux-only tools).
If you are using for server for other things too, you might want to look into a monitoring system with grafana, prometheus and node exporter. It records a lot of different stats, most of which you won’t use, but cpu stats including iowait, memory stats including used and cached, and the like can be very useful in these situations. Sometimes there’s a bottleneck that’s not obvious without seeing how a stat was changing over a longer time or out of context, and depending on what monitoring too have you used so far, it might have not shown you important things like aspects of disk utilization. Enterprise grade drives can also have poor performance in certain situations.
I’m honestly sorry for that instead of everyone else. I think it’s just that we haven’t experienced problems like these, and you were a bit harsh, in the sense that you are confidently saying it’s Jellyfin’s fault, where it is probably something else.
If you feel like it, you could keep testing it “in secret” (not from us, from your wife), and when you have fixed it and been going stable for a week or two announce to her that you found the problem, it was on you (if it was), and now it’s much better, and reintroduce it to “production” use.
Though it’s a bit weird that besides perf issues, if I understand right you also have major bugs. If you only tried it with the chromecast, test it with normal jellyfin clients (web client, desktop client, mpv shim, tv app, whatever you like), and see if it’s better there. I suspect it has to do with the chromecast.