After switching to Plex Pass PMS is constantly utilizing one core (version 1.14.1.5488)

Plex Media Server NAS & Devices
46

Yes, I did, it was my OpenWRT router. Can you provide any further detail on what the incompatibility with the XML spec may be and I’ll feed it back to the OpenWRT forums where I’m active. I’m happy to tcpdump a conversation if that’s of any help. I’ll have a look at the miniupnpd spec and see if I can figure it out.

But why would Plex even be attempting to talk to a UPnP daemon if I’d told it I was manually configuring port forwarding? Expected behaviour if PMS were badly interacting with another system would be to tell PMS to stop interacting with that system (ie. manually configure port forwarding) and leave that other system alone for the services it’s working correctly with. Having to disable that service, and have it no longer provide its other clients, is not great, imho.

47

Okay, a little more experimentation.

  • As soon as I start miniupnpd with only UPnP mode enabled on port 5000 PMS goes 100% CPU (even though it’s not configured to use UPnP for remote access)

  • Running miniupnpd with NAT-PMP enabled but no UPnP, does not produce the high CPU behaviour

NB any changes to miniupnpd configuration when PMS is already in a high CPU utilisation state requires a restart of PMS to bring it back down. Conversely, if PMS is operating normally, with CPU idling, it will go into a high CPU utilisation state as soon as UPnP services appear on the network.

1 Like
48

@gary_parker that worked!

In OpenWRT/LEDE, I disabled “Enable UPnP functionality” under UPnP, while leaving “Enable NAT-PMP functionality” enabled.

I then restarted my Plex docker container, et voilà! CPU usage is back to normal. It’s now using a little less than 2%, instead of close to 25% CPU.

Thank you!

EDIT : I can also confirm that this changes the NAT Type on my Xbox One from Open to Moderate, as well as saying “UPnP not successful”.

2 Likes
49

Thanks for the feedback @SX86 and happy it’s working better for you now! Also, interesting to see that you’re running OpenWRT, too. I wonder if we can narrow this down to a specific interaction between the latest versions of PMS and the version of miniupnpd that’s running on OpenWRT (miniupnpd - 2.1.20180706-1).

miniupnpd (https://github.com/miniupnp/miniupnp/releases) doesn’t appear to have had any active development since May '18, so it’s unlikely to be a change in the UPnP service that’s caused this behaviour, it’s more likely to be a change in the PMS code, imho, especially as we can mitigate by rolling back to older versions of PMS.

edit just had a user in another thread confirm that turning off UPnP on his router (Amplifi HD, very likely based on OpenWRT) stoped the high CPU issues with 5488.

2 Likes
50

I just tested this myself: I have OpenWRT and miniupnpd installed on my router; with UPnP enabled Plex goes into some sort of a loop and consumes 100% of a core on my server. Disabling UPnP in miniupnpd’s settings solves the issue. Leaving NAT-PMP enabled or disabled has no effect.

3 Likes
51

That’s great news, thanks for letting me know @gaygirlie_hotmail.com. Hopefully now @ChuckPa and the Plex team can find an OpenWRT router in the office and start investigating the problem with the 5488 and onwards builds.

1 Like
52

Hey @ChuckPa, this looks like the thread where UPnP was first identified as the problem for 1.14.1.5488 and 1.15.x builds of PMS. The author did some decent debugging on a Win10 machine to discover it: Another 100% CPU on one thread post: cppnetlib-uri.dll!boost::network::url::detail::parse [solved]

1 Like
53

Are you also using DHCP with your configuration?

The most stable, and always advocated, is to use a static IP for the PMS server. Doing so allows manual ports to be opened if necessary.

54

I’m using a static IP, in my case. DHCP is enabled for other devices, but not my Plex Servers.

55

Are any of the other devices a router / IGD with UPNP enabled on it?

56

The server I’m running Plex on has a static IP address. Most other devices on the network are DHCP with static assignments. I have no other devices on my LAN acting as a router or with UPnP daemons enabled (although, obviously, there are lots of other UPnP enabled devices).

57

Using DHCP or static IP doesn’t matter, at least in my case, and neither does using manual port-forwarding or UPnP – as long as UPnP is enabled at all on the router, PMS goes all into a loop. Also, no, I do not have any other routers with UPnP on my network.

58

Thanks, Conducting tests now and working with engineering.

What make/model modem/router do you have. I ask because most have UPNP active on their routers. A major flaw would have long ago been discovered.

59

@ChuckPa
I’m using a :
Linksys WRT1900AC v1,
running OpenWRT 18.06.1,
and miniupnpd 2.1-1.

1 Like
60

I think you just nailed it. “miniupnpd” . I have heard reports of that not being adequate.
I will speak with Engineering.

Do you have an alternative UPnP daemon module you can run? Something in the same calibre as pfSense?

61

while I look,

option enable_natpmp 1

is also set ?

62

@ChuckPa Yes, NAT-PMP is enabled.

Thanks for looking into this!

EDIT : I haven’t looked into other UPnP daemons for OpenWRT… not sure if there are any alternatives.

63

Apologies if I am cluttering the thread, but just wanted to +1 the symptoms and remedy.
With plexpass, running PMS on an unraid server using an OpenWRT router with miniupnp I get 100% usage of one core on the plex server and the UDP socket count spirals through the roof.

The instant I disable the miniupnp daemon on the router, the cpu usage on pms drops to 0%.

I also have not heard of anyone else using any other upnp daemon on OpenWRT than miniupnp.

Some good debugging in this thread to nail down the issue, well done guys :slight_smile:

1 Like
64

Data collected and submitted to Engineering.

Referencing other thread.

65

If theres anything we can do on our OpenWRT routers to provide testing or data for you just let us know.