PMS is "killing" my synology (CPU+RAM) since intro detection on tv series feature

[FlexiPlexi]

Server Version#: 1.19.4.2865 (I tried the lastes beta, but that did not help)
Player Version#: win/mac 1.11.0.125-1d8c460c
Synology: DS1815+ with DSM 6.2.3-25426

Hi all,

since the introducing of the “Skip TV Show Intros” feature, PMS is overloading my DS1815+
I tried to disable that feature in the server settings, but that did not help.

After a while CPU and RAM is occupied by PMS:

It takes about 30 mins to connect to console or open the package manager from the synos webgui, because PMS is taking all ressources.

If I’m able to stop the PMS process, everything is going down to normal.

I tried downgrading the “normal” way, but that’s not possible within the package-manager and I currently don’t have the time to do it by hand.

Is this a known issue?

[FlexiPlexi]

@trumpy81

but I suspect you have some damaged/poorly encoded files among your media collection that are causing Plex to loop and not complete each thread properly.

As I indireclty said before, PMS did not overload the NAS until the new feature “skip intro” was introduced.

Regarding your above suspicion, I guess that means, PMS has now changed its behaviour to scan mp3 files via ffmpeg, which I understand, you are refering to.

Thanks for your thoughts, I will remove the series folder from the NAS, and let PMS rescan the music and series libraries.

If PMS is not consuming all of the NAS’ RAM+CPU, we’re surely have to check the music folders to find something to blame for, but if it is working “normal”, there is no need to hunt something in the wrong place… (series or music).

I will try to prepare some logs though.

[FlexiPlexi]

@trumpy81

Where did I mention mp3 files?

Thank you very much for pointing that out! I must have been in a state of delusion and misread/misunderstand your answer. Sorry for the confusion and the waste of time!

I’m on the way to collect the logs, it’ll surely take some time.

[FlexiPlexi]

@trumpy81

I am in a constant state of delusion

Sometimes (but more and more often the last years), I really think, that could (or should) be the way to spend the rest of my life… in a nice and happy state…enjoying all the things life offers instead of trying to ignore (or even fix) the parts I don’t like. Until now, I did not come to a conclusion which I can not ignore … seems like a vicious circle…

@trumpy81

I strongly recommend you run your files through ffmpeg searching for errors and if any errors are found in your files you should re-encode or replace them.

Is there a suitable command (console), or maybe a program with a gui, which I can run over the folder containing all the series, or how would I do this?
Or a nice support article, which I’m not able to find?

I’m able to use the linux command line as well as a mac or windows tool.

[FlexiPlexi]

I don’t

it takes some time/hours to scan the libraries and after the scan is done, PMS is analyzing the single episodes for intros. Then, after some time again, PMS begins to overload the NAS.

It seems, PMS is not analyzing episodes w/o scanning the series-libraries first, so we have to be patient.

I guess, I can send you logs tomorrow.

[ChuckPa]

If I may augment here?

I also own a DS1815+.

The Atom C2538 CPU is not the fastest tool in the arsenal.

There is also loudness analysis happening at the same time.

Please use regular top when reporting Synology CPU loading.
htop , especially in that view mode, is useless as a diagnostic. We need to see pids, not threads.

I will conduct my own testing.

[FlexiPlexi]

Hi Chuck!

@ChuckPa

Please use regular top when reporting Synology CPU loading.

Will do next time.

Meanwhile, I adapted trumpy81’s windows batch to a bash shell,
which unfortunately logs a lot of would-like-errors as noticed by trumpy81.

@trumpy81

Do note, that not all errors are hard errors. Some errors seem to have no effect on playback, while others can be troublesome, but not in all situations.

I’m not sure, how to decide, which is a critical error and which not…

Anyways, here is the script, maybe it’ll help someone out here:

#!/bin/bash
# This script uses FFMPEG to find possible damaged video files.
# Logs found files in a given directory (WHERE2SCAN) to a logfile (LOGNAME) and protocolls its errors.

WHERE2SCAN="/volume1/public/videos/"
LOGNAME="/volume1/public/PlexVideoCheck.log"
shopt -s nullglob nocaseglob

echo `date` "### START" >$LOGNAME

find "$WHERE2SCAN" -type f \( -iname "*.MKV" -o -iname "*.MP4" -o -iname "*.AVI" \) -exec echo `date` "### CHECK " >>$LOGNAME {} \; -exec bash -c 'ffmpeg -v error -i "$0" -f null - 2>>$LOGNAME' {} \;

echo `date` "### END" >>$LOGNAME

@ChuckPa:

This is a somewhat personal matter for me, you know it,

I really cannot agree to your argument of CPU power might be a reason here.

It should be up to the developers to check what power the underlaying hardware is capable of and reduce the work of the softwares to an amount, so as less RAM and CPU is used as possibe while giving a nice and smooth (aka fast) experience to the user. (Remember the times, where 640kb was enough for everybody… )

While this might not an easy part for the developer, (I see that)
it is surely not the right way to overload the users machine or just do not free the memory which is not needed anymore.

The loudness analysis uses CPU+RAM too, but works like a charme. There are surely defective broken music files too on some users collection, but here PMS seems to handle those broken files correctly without overloading the hardware.

IMHO it is a NO GO to ask the user to re-encode/delete/replace/etc files, just because the underlying code does not care about the garbage it produces.

In general, my mantra is:
The application (PMS) is responsible for overloading the machine, not the maybe broken files on the users harddrive.

If the application can not handle broken files (which is totally acceptable), then why is it not just skipping those broken files? As simple as that. Skip everything, which is not readable. Tell the user in the logs. Continue the work.

Fixing the PMS code is the only correct way…

Having your users re-encode or delete crappy files or whatever, will just produce unnecessary rat tails of problems.
(Of course, if files are crappy, the user should replace them in the first place, but not because PMS cannot handle those and overloads the users machine)

It seems I’m from the dinosaur age, where a good and working code was necessary to sell an application in the first place.

Enuff for tuday!

@ChuckPa

I will conduct my own testing.

Thank you! Let me know, if I can asist you in any way!

[FlexiPlexi]

@trumpy81

Attach the Logs.zip file to your post, thanks.

Finally the logs:
[Logs.zip|attachment](7.7 MB)

A “top” screenshot, when I stopped the plex package. Don’t know, if it is useful to @ChuckPa:

@trumpy81

Why do people always cherry pick my statements?

Have you ever experienced a similar situation, like:

Someone asks you for the way and you tell him: “yeah, I know the way, its easy, just go straight ahead, then, wenn you see a traffic light, turn right, go 30 meters to the next crossing, turn right, then left again and go further until you see a big white house…”

Or in the supermarket, when you ask an employee for a special soup: “its the shelf next to the loundry-shelf, just go behind that floor then you see it.”

We’re all caught in your own head, if you understand those examples

In addition to that, I surely can not not re-encode most of the video files (again, because I handbraked most of them a long time ago), just because of the sloppy and rude PMS code until we find the real responsible for the overload of the CPU+RAM.

Hope you can find anything useful & related in the logs!
(I will remove them later on, because they contain sensible data).

[FlexiPlexi]

Holidays are over

Can anybody of the support open a private conversation or a secure channel for the logs?

I’m not feeling good posting logs here, which contain email-adresses of the users connecting to the PMS.

Thank you.

[ChuckPa]

To confirm the problem really is Intro Detection:

1. Settings - Server - Library - Show Advanced
2. Set Generate intro video markers to never
3. and save the setting
4. Restart PMS

[FlexiPlexi]

@ChuckPa

• Settings - Server - Library - Show Advanced
• Set Generate intro video markers to never
• and save the setting
• Restart PMS

Confirmed, did that last week.

@FlexiPlexi

I tried to disable that feature in the server settings, but that did not help.

[ChuckPa]

Take the next step.

1. Stop Plex
2. Uninstall PMS
3. Re-install the previous SPK which you downloaded and installed.
4. Confirm normal operation.

[FlexiPlexi]

@ChuckPa

Re-install the previous SPK which you downloaded and installed.

How would I do that?

@FlexiPlexi

I tried downgrading the “normal” way, but that’s not possible within the package-manager and I currently don’t have the time to do it by hand.

[ChuckPa]

Kindly read my instructions: Steps 1-4 above. They tell you precisely how to do it.

Did you not “Manual Install” PMS which is what got you to this point?

[FlexiPlexi]

@ChuckPa

Uninstall PMS

Won’t that remove all of the database and the changes I did the last months?

[ChuckPa]

No.

I keep the Plex share separate from the software - deliberately.

Reasons are multi-fold:

1. The Plex share persists when the package is removed
2. The package itself, in /volumeX/@appstore would be removed.
3. I do not wish to subject users to hours of rescanning because synology does not allow package downgrade in Package Center.

[FlexiPlexi]

@ChuckPa

No.

Great!

@ChuckPa

I keep the Plex share separate from the software - deliberately.

I installed/updated Plex via Package manager, so I didn’t separate anything “deliberately” but as I understand you, I just can uninstall Plex via Package manager and the whole Database, Caches, Plugins, etc (living in /volume1/Plex) stay untouched and only stuff in /volume1/@appstore are removed.
Also my settings are save.

Sound perfect, thank you.

I will uninstall/reinstall Plex and give feedback, once it runs 2 days without overloading.

[FlexiPlexi]

Downgrading PMS (to v1.19.3.2740-add6f438d) did not seem to work.
PMS is crashing every night.
Thats worse too…