Plex Server (Synology DS-3615xs): Playback does not start (stuck on playback spinner)

[G3orgios]

On Plex Server (Synology DS-3615xs) the playback does not start and gets stuck on playback spinner for ever. Troubleshooting done, with PMS restart after each action, with no success:

1. Update to latest PMS ( 1.19.2.2737)
2. Tried Smart-TV and Web-player
3. Configure manually the transcoding directory to /volume1/Plex/tmp_transcoding/ (long time ago this worked and solved the “no-playback-starting” problem. Not this time…)
4. Clean-up the codecs (.so files from /volume1/Plex/Library/Application Support/Plex Media Server/Codecs/99c90e0-3095-linux-x86_64 removed for PMS to fetch them anew)

After may tried and PMS restart no luck! Logs are attached.

Thank you.

Server Version#: 1.19.2.2737
Player Version#: latest Firefox / Smart-TV / Android
Logs: attached (NO verbose logging!) Plex Media Server Logs_2020-05-04_10-06-13.zip (6.1 MB)

[G3orgios]

So, problem SOLVED and thank-you go to @trumpy81 for pointing to the right directions to look into! Some troubleshooting details for any future reference:

1. Network Adapters: I use a 10Gbit Intel network card and all Synology network adapters (4) were disconnected (i.e. no physical connection, although in the CP you can see the last configured IP address for each one, which cannot be removed). Strangely, it seems that PMS keeps remembering older used IP address (…probably to try them out in case of no connectivity?). Followed the suggestion of @trumpy81 to deactivate IPv6 plus configured network on my primary adapter (port-1 of 10Gbit Intel card)
2. User rights: confirmed “plex” user had full (admin) rights on temp trans-coding folder
3. Trans-coding: …tough to have been a faulty/failing disk since no health issue was reported by Synology on both Volume and Disk level, running also on RAID-6. However, I gave it a try and switched trans-coding temp location from volume1 to volume2. Strangely, seeing the console logs, Plex continued trying to trans-code to volume1, even after service restart!

After every single (1) change PMS was restarted but all these produced no results. ONLY after making a full server restart (reboot) trans-coding came back to life! Then, changed back trans-coding temp folder to default (i.e. no location path manually defined) and transcoding was again on-line.

Bottom-line, it seems that for some reason PMS got some network hick-ups that were rectified by deactivating IPv6 and/or defining a specif network interface as preferred.

Thank you.

[G3orgios]

Sadly the Network parameters trimming provided a resolution only shortly.

I was getting quite many “EAE timeout! EAE not running, or wrong folder?” errors (console view) and looked it up. I have quite a few, and then some, folders of movies/tv-series/etc, thus I am now giving it a try with “Increase the number of directories Linux can monitor (Notify)” by adding fs.inotify.max_user_watches=262144 in the /etc/sysctl.conf.

Let’s how it will go and i will update hereby.

[G3orgios]

I guess, as @trumpy81 has rightly pointed out, there are quite a few different issues underlying here. I am not sure that increasing the number of monitored directories did much however I have set it to fs.inotify.max_user_watches=65535 since I have 35K+ folders.

On the Network aspect, Synology DS-3615xs has 4 Ethernet adapters plus i have a 10Gbit SFP+ card with 2 more. The four (4) default Ethernet adapters of Synology are deactivated (i.e. no cable connected, with internally assigned IP: 169.254.x.x) but is really strange that PMS is still browsing on ALL six (6) of them, including the Four (4) deactivated ones! Why so?!

Even more strange is that when PMS is active (i.e. streaming content) there are few warnings/errors BUT when PMS is idle there is a flood (e.g. “Network Service: Error in browser handle read: 125 (Operation canceled) socket=-1”, browsing the deactivated Ethernet ports, etc). Possibly that’s why either:

1. PMS gets eventually stuck on about 12% CPU usage (on idle!) OR
2. Trans-coding functionality ceases and when trying to playback something it gets stuck on playback spinner?)
   NAT: PMP, got an error: Not Supported by gateway.

Any ideas/suggestions?

[ChuckPa]

@G3orgios

I was getting quite many “EAE timeout! EAE not running, or wrong folder?” errors (console view) and looked it up. I have quite a few, and then some, folders of movies/tv-series/etc, thus I am now giving it a try with “Increase the number of directories Linux can monitor (Notify)” by adding fs.inotify.max_user_watches=262144 in the /etc/sysctl.conf.

On Synology, this isn’t how we increase the number of directories PMS can monitor. We used to do that but Synology didn’t like us changing it so they obfuscated it.

HOWEVER

It’s best to verify we need to increase the table prior to just asserting a new size.
Are there “No space left on device” errors while Notify is initializing ?

If we do need increase, we need to do it this way

Synology FAQ - Questions, Answers, and a few How-To's

FAQ 16: Increase the number of folders Plex can monitor Applicable to all supported versions of DSM. When the Linux ‘inotify’ table fills up, this will cause when Plex to stop automatically detecting new media you add to the server. It will also manifest when there are transcoder problems with Dolby audio codecs. Linux has a default table size which is static. We can increase this at will. On some versions it’s 8192. On newer kernels, the default is 65536 folders. This is easy to in…

Then restart DSM.

Per the CPU stuck, I’d like to see some logs please; DEBUG logs ZIP; gathered about 2 minutes after PMS starts and having sat as idle as possible during that time.

[G3orgios]

Thank you for the feedback @ChuckPa! With an idle PMS service, the service process goes up to even 80-90% CPU usage, and I run a DS-3615xs on an upgraded CPU (Xeon E3-1285L v3 3.1GHz) and RAM (16GB).

The folders table increase is a must since there is a flood of such errors: [Notify] Failed to add watch for “/volume2/Videos/TV Series/xxxxxxxxxxx” (28: No space left on device)

I will give it a try, with the right Syno-way this time, and get back.

[G3orgios]

No luck with changing the folders table. Meaning I did the math as per Synology FAQ for my folders (45K) and started with s2s_watches_max=65,536. Still i was getting (in console) a flood of Notify alerts and started making gradual increments up to 2,147,483,648. Synology was emailing me every time the task was running successfully and was also restarted after every change. Still, Notify alerts persist!

@ChuckPa DS-3615xs has FCLGA1150 CPU socket and the original i3-4130 was replaced with a Xeon E3-1285L v3. PMS is running natively on the machine, no docker/XPenology/other “middle-ware”. And here are the logs of the first few minutes after reboot of Synology DS-3615xs and with PMS idle (no streaming or other user activity): Synology DS-3615xs_Plex Media Server Logs_2020-05-08_21-07-51.zip (6.5 MB) Please have a look and let me have your feedback.

[ChuckPa]

If you would please do the following

1. Restart Plex
2. Wait 1 minute for it to stabilize
3. Attempt to open it
4. Wait 30 seconds
5. If no success, stop here
6. If success, continue until you get “stuck”
7. Once stuck, Stop Plex
8. Control Panel -> Shared Folders - EDIT the Plex share -> Permissions tab
9. Give your username R/W permission
10. File Station - Plex - Library - Application Support - Plex Media Server
11. Right-click Logs
12. Compress to Logs.zip
13. When complete, click to download to your computer
14. Upload here for me to inspect please.

[G3orgios]

@ChuckPa PMS is always starting properly, is accessible and responsive via Web-UI BUT the problem is that is does not start payback (trans-coding), even after full server reboot, and that sometimes, in-time (hours/days), the process itself might get stuck. When this happens, the CPU is stuck on a 12-14% load (as if PMS is streaming normally, although it doesn’t) or at worst the CPU load is gradually increasing to almost max-out the CPU (even up to 90%). In the previous post you may find the logs retrieved via PMS web interface after a full server reboot and for a few minutes with the PMS idle (no streaming, not even web access to UI). I hope you might dig out something!

Regarding the latest procedure to retrieve the logs manually, I will leave PMS aside to eventually get into the no-reason-CPU-load status and get you the logs. Thanks.

[G3orgios]

@ChuckPa here we go, PMS is on 12% CPU usage, supposedly as trans-coding, although it is idle with no clients connecting or performing any other user task (folders scan, etc)

Here are the logs, retrieved via the web-UI, since PMS’ web-UI is fully functional!

Plex Media Server Logs_2020-05-09_14-48-33.zip (6.1 MB)

[G3orgios]

@trumpy81

May 09, 2020 12:27:02.758 [0x7fb69eb2e700] WARN - HTTP error requesting GET https://185-36-142-96.b21104a318494a57a9db40fbd7fbf5b9.plex.direct:26774/identity (7, Couldn't connect to server) (Failed to connect to 185-36-142-96.b21104a318494a57a9db40fbd7fbf5b9.plex.direct port 26774: Connection refused)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] ERROR - getaddrinfo(192-168-123-2.abcdefghijklmnopqrstuvwxyz012345.plex.direct) failed: -3
May 09, 2020 12:27:07.040 [0x7fb69df76700] ERROR - PublicAddressManager: Unable to get public IP adddress from myPlex (httpCode=-7): 

Plex is unable to communicate properly with plex.tv on your network.

@ChuckPa I have raised this point earlier, why is it that PMS is trying to communicate with via IP address which is not valid? I had PMS installed and running on Synology under 192.168.123.2 but long time ago i moved to 10Gbit adapter with other IP (192.168.123.12) and any older IPs used are not valid (i.e. adapters have no electrical connection to the LAN). Why is PMS still trying to have any communication via a disconnected port and not active IP?

@trumpy81

If your router has any DNS rebinding support, then make sure you allow the *plex.direct domain to have full access.

Added on my Mikrotik

@trumpy81

I would also advise you to change the name of your volume2/Videos folder to something else (Media perhaps). Synology reserves the video folders for Video Station, and that has been known to cause issues with Plex.

Video station was never installed/used on my Syno.

Going to reboot and will monitor and report here. Thanks

[G3orgios]

@ChuckPa After the reboot the trans-coding starts properly but let’s see if it will crush later on…

I Did the DNS rebind to allow PMS IP address (192.168.123.12) BUT PMS is still trying to get communication on the non-existing 192.168.123.2! (the primary IP used when I installed PMS some years ago)

ERROR via console: “getaddrinfo(192-168-123-2.abcdefghijklmnopqrstuvwxyz012345.plex.direct) failed: -3”

Why so? How can we get PMs to “forget” older IPs? (of course, I have the proper IP 192.168.123.12 selected as “Preferred network interface” on PMS)

[ChuckPa]

It’s possible to get unused (unplugged) adapters to come up with inactive IP addresses

1. plug in a wire
2. Set to DHCP
3. Let it pull an IP
4. Un-plug the wire
5. Complete for other adapters
6. Restart DSM

When restarted

1. Verify DSM has the 1 active adapter checked as the Default Gateway
2. Go to PMS (Settings - Server - Network - Show Advanced)
3. Select the same preferred adapter.

For future reference.

Placing two adapters on the same subnet is not wise. It confuses PMS because DSM uses the two adapters interchangeably and indiscriminately . Extreme care, executed at the command line level, must be used in such cases.

I used this as reference

May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 1 lo (127.0.0.1) (loopback: 1)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 3 eth0 (192.168.123.2) (loopback: 0)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 4 eth1 (192.168.124.2) (loopback: 0)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 5 eth2 (169.254.70.68) (loopback: 0)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 6 eth3 (169.254.250.162) (loopback: 0)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 7 eth4 (192.168.123.12) (loopback: 0)
May 09, 2020 12:27:07.032 [0x7fb69d6ac700] DEBUG -  * 8 eth5 (192.168.124.12) (loopback: 0)

[ChuckPa]

How was DNS rebinding resolved for the LAN?

Is there a DNS resolver / server on your LAN?

[G3orgios]

@ChuckPa

How was DNS rebinding resolved for the LAN?

So far so good! It seems that no hic-ups on trans-coding since i configured this.

@ChuckPa

Is there a DNS resolver / server on your LAN?

I am not sure what is the DNS resolver but DNS resolution on my Mikrotik is configured to external DNS (not on LAN) using three WAN DNSs (primary, secondary, tertiary)

@ChuckPa

Placing two adapters on the same subnet is not wise. It confuses PMS because DSM uses the two adapters interchangeably and indiscriminately . Extreme care, executed at the command line level, must be used in such cases.

I have only two (2) eth-ports active (one 10Gbit expansion card on synology) and they are on different class-C networks. One is the primary one that goes to the managed switch (WAN access) and the other is only to connect with a 2nd Synology (backups).

@ChuckPa

It’s possible to get unused (unplugged) adapters to come up with inactive IP addresses

1. plug in a wire
2. Set to DHCP
3. Let it pull an IP
4. Un-plug the wire
5. Complete for other adapters
6. Restart DSM

When restarted

1. Verify DSM has the 1 active adapter checked as the Default Gateway
2. Go to PMS (Settings - Server - Network - Show Advanced)
3. Select the same preferred adapter.

By inactive IP do you mean 169.254.x.x? I have done the DHCP trick (without an electrical connection) but i still see PMS reporting these “inactive IPs”. I will check it again

@G3orgios

No luck with changing the folders table. Meaning I did the math as per Synology FAQ for my folders (45K) and started with s2s_watches_max=65,536. Still i was getting (in console) a flood of Notify alerts and started making gradual increments up to 2,147,483,648. Synology was emailing me every time the task was running successfully and was also restarted after every change. Still, Notify alerts persist!

Any idea why I am still getting these Notify alerts? Despite pushing the folders table size to the roof?!

[ChuckPa]

May I see what you actually added?

This would be incorrect syntax.

s2s_watches_max=65,536

This would be correct

s2s_watches_max=65536

[G3orgios]

@G3orgios

s2s_watches_max=65,536

Typo mistake in my post. I added s2s_watches_max=65536 Then i stretched it gradually to much higher values (without any number-separator) up to

echo ‘s2s_watches_max=2147483648’ >> /etc/synoinfo.conf

but still getting a flood of Notify alerts.

[ChuckPa]

Did you restart the Syno?

FYI: That value will be rejected. Each watch requires 540 bytes of memory.
Do you have 10 GB of memory to waste?

You can’t keep appending to /etc/synoinfo.conf like that. It will corrupt DSM.

[ChuckPa]

Please use this procedure. It works.

Synology FAQ - Questions, Answers, and a few How-To's

FAQ 16: Increase the number of folders Plex can monitor Applicable to all supported versions of DSM. When the Linux ‘inotify’ table fills up, this will cause when Plex to stop automatically detecting new media you add to the server. It will also manifest when there are transcoder problems with Dolby audio codecs. Linux has a default table size which is static. We can increase this at will. On some versions it’s 8192. On newer kernels, the default is 65536 folders. This is easy to in…

[G3orgios]

Yes, I used the Synology FAQ article you had sent me and I had restarted after every incremental change but still getting Notify errors. I have about 45K folder. Shall i stick to s2s_watches_max=65536 ? Maybe better to deactivate the DB monitoring (I don;t mind to do manual updates when i add content)

Still getting continuously “Network Service: Error in browser handle read: 125 (Operation canceled) socket=-1”

You mentioned reinstalling DSM. I would mind making a fresh install IF I don’t lose the content and the PMS DB (edits done, watched status, etc). Is this feasible?