I’ve been using it for the past two years without problems and just now I did.
Is there maybe a specific requirement that changed?, I can always ask them to install it.
I’d love to get support instead of a closed answer.
You have been fortunate.
The specific details are:
- Feral hosting uses Linux Namespaces to isolate their customers from each other
- This means all processes co-exist in the same process table versus the containers / VMs other providers utilize
- When PMS starts, since it only allows one instance of itself per host, makes certain it doesn’t have and hung processes. Upon seeing any of its processes, it terminates them. This is the initial incompatibility with Feral. No two customers may use PMS concurrently on their hosts.
- The next incompability is their choice of runtime libraries. They favor the older versions. With 1.15.x, PMS’s libraries have been updated. Of most importance is glibc to 2.14.1.
- They favor Python 3 and inject it into the PATH and LD_CONFIG in front of PMS. PMS requires Python 2 (shipped with PMS).
Right now PMS is running with Python2 (and I’m completely sure of that), but the single process instance might be a killer.
Was this introduced specifically in 1.15 or it was always like that? Because if it was like that since forever then the issue is probably not that, otherwise it would’ve broken before.
On another hand, the glibc update might be it and that can be updated per user request.
I’ll wait for an answer before requesting anything.
I can assist you if you would please create a VM (not a docker) and recreate the issue.
Feral hosting’s servers are so non-standard that I cannot support them. We do not have a feral account. Their servers are equivalent to using an unsupported Linux distribution and getting unexpected results.
As I first stated, you have been lucky.
Can you run docker on feral? If so it’s probably easiest to migrate to that. It was fairly easy to do so and I haven’t looked back.
I can’t. But they do allow users to request the installation of different dependencies.
Can you please answer my questions at least before you just blow me away and say “it’s not possible”?
Was this single process per instance a change that was introduced on 1.15 or has it always been like that? How would you have treated this issue if the word “feralhosting” wouldn’t have come up at all? Maybe this is a totally unrelated issue to feralhosting in particular and could be easily solvable through other means.
I’m not asking you to solve my issue on your own, I’m not even asking YOU in particular to solve it, I don’t assume you’re some kind of miracle helper, I’m a dev as well and I think I can figure some ■■■■ out on my own as well, I just need some questions answered in order to maybe find the root of the cause.
Honestly, the whole attitude of your support sounds like you don’t even want to help, but instead just wanna blow me away and do no support at all. That’s very annoying and I believe that’s not the point.
Plex has always been structured to be one server instance per host.
There have been Feral customers who had the luck of being the only Plex user on that particular host.
Ok, so I can assume this problem has nothing to do with the fact of the single process per instance or Python2, maybe it is the glibc that’s not updated, I will check that and post my findings.
Thanks
As supplemental information, PMS 1.15.0 raised the glibc() minimum version required to 2.14.1.
Previously compatible distro/versions now won’t work. Example: Debian 6
Cool, thanks!
Before anyone gets all up in my face about it,
Plex working on Feral hosting is “luck”.
It was never tested
It isn’t listed as officially supported.
If it works, great. If it doesn’t and nothing can be done about that, then it’s unfortunate.
This leaves Feral in the same classification as Arch, Genaro, Gentoo, Mint, and other distros.
Well, after digging some and then more. The problem is not really about feral or it’s customized version of the kernel.
Specifically what you pointed out about their practices is not quite so:
About 1, 2 and 3.
They do isolate namespaces, but currently the process table of other users is completely hidden from one another so that doesn’t interfere with the “single process per instance” issue.
Even more, the error was never about the process finishing early or being killed like you mentioned, but instead it was all about PMS not being able to load the included module “unicode.so”, which wasn’t loaded because there was a problem with finding the specific _PyUnicodeUCS2_ToDigit symbol.
-
Feralhosting is already running with a newer version of glibc than 2.14.1, it’s using 2.24 (2.24-11+deb9u3) and this version is not customized nor modified in any way by them.
-
They are not injecting python3 into the path or the LD_CONFIG any more than the standard python3 installation already does, this I found by inspecting the screen in which plex is running. I did change the path and ld_config to accept that of python2 and ld 2.24 just to be double sure (even though before they were already correctly placed) and the problem persists. Even if this was misplaced, like you said, Plex ships it’s own version of Python and in any case it’s clear in the logs that in fact that binary is being used.
I haven’t yet found a solution for the symbol problem, my guess is that this has something to do with glibc but clearly not just its version. I’m still digging, but I just wanted to clear out some of the possible issues mentioned before.
So it appears they have made changes this past year. Thank you for that.
This, however, does not change their status as a non-supported distro.
One cannot deny they utilize their own kernel build. By using their own kernel build, one must ask what other changes have been made, further reinforcing that it will remain a non-supported platform until such time there is sufficient demand to do so.
There is low probability of officially supporting Feral Hosting. Doing so would show a bias / endorsement which does not exist.
Yeah I understand, I’m not really asking for support for feralhosting, but I thought there might be others out there that might run into the same issue.
Especially when running a docker instance instead should likely solve the problem at hand, as it did with my Gentoo issues. It was fairly straight forward, and took me a fraction of the time to set up compared to the time I spent trying to figure it all out, with Chuck’s assistance. Thank you Chuck by the way, again.
I understand running docker with a mounted volume can be an easy fix, but:
-
There’s no services that support the same specs as Feralhosting for the same price (I pay roughly 20GBP for 2TB, shared 24 cores and 126gb of ram, an INSANE internet connection and ssh access)
-
I consider feral to be a rather good service, easy to work with, great customer support and very flexible for other use cases rather than just Plex. I also run Sonarr, Radarr, Rtorrent and Jacket inside it. In a per-container hosting this would be much more expensive even without the volume and for a cloud VPS provider this is WAY more expensive.
-
I just like debugging stuff like this, I consider it a little pet project so I don’t mind the troubleshooting.
-
I don’t wanna have a homelab type of setup.
Disclaimer: I don’t mean to make this a shameless plug (I’m in no way associated with Feralhosting besides being a client) and I 100% don’t expect anybody in Plex giving me more than basic support for my troubles.
I wasn’t implying to move from feral. I was making the assumption if you could run plex binary, you could also run docker?
Can’t run docker in Feral… not at the time of this post at least.
Did you find solution yet? I’m having same problem, but on different OS. I’m getting _PyUnicodeUCS2_ToDigit error while scanning, and not updating library…
