Hi @ChuckPa, also running OpenWRT (custom build based on 18.06) with miniupnpd - 2.1.20180706-1 on a Linksys WRT32X here. NAT-PMP is enabled and UPnP is now disabled. Minupnpd is the standard UPnP daemon across all openwrt builds and has been for many years. There are a lot of commercial router firmwares based on OpenWRT so many users will be running it and perhaps not realising.
This behaviour in PMS when miniupnd is active on the LAN with UPnP enabled (not NAT-PMP) only appeared with the 1.14.1.5488 and 1.15.x builds. Itās reproducible on those builds and goes away when 1.14.5470 is installed instead.
Last bit of additional info from me for a while, hopefully: feedback from the OpenWrt forums indicate that at least the Turris Omnia and Gl-Inet commercial routers are also based on OpenWrt (in addition to a number of Linkysys and Ubiquiti devices and all the derivative firmwares such as Gargoyle, Tomato, DD-WRT, etc.).
As before, Iām more than happy to carry out testing or provide more info if engineering want it. Iām keen to see this get fixed as Iāve since discovered that disabling UPnP broke Remote Play on my PS4 (ports now manually forwarded).
Interesting to see that itās not limited to the OpenWrt build of miniupnpd, and also affecting non-OpenWrt installs. miniupnpd is, obviously, the common factor here.
Iām really curious to know what PMS started doing differently to get so upset by this particular UPnP implementation in itās most recent releases as itās worked flawlessly with miniupnpd for years now.
I can confirm this behaviour is still present in the beta release 1.15.0.647 (and so see no reason itāll be fixed in the production release):
upgrade to 1.15.0.647
turned on UPnP
saw Plex Media Server process go to 100% CPU within a second or two
turned off UPnP, no change in CPU usage
restarted Plex, CPU usage returned to normal
Plex 1.15.0.647 running on XPEnology server (based on Synology DSM 6.2-23739 Update 2) with a Linksys WRT32X running davidc502ās custom OpenWrt build (r9028). LAN is NATād IPv4 and routed IPv6 /64.
Iāve hit this in 1.14.1.5488 and 1.15.0.647. Iāve been keeping old self-built 1.14.0.5470 packages around to work around this for a long time now. Itās an Arch Linux host acting as a router and server, running miniupnpd v2.1.20180706. I can confirm the same behavior @gary_parker mentions above.
Iāve restarted minupnpd a number of times now, with UPnP turned on and off and also restarted the Plex service and, beyond the first time after upgrade, itās never gone 100% CPU. I think this may be the fix!
edit
Gah, spoke too soon. Rebooted my router and PMS is back to its old tricks. Sorry
No Arch Linux here. Im using linuxserver.io docker on Unraid 6.6.6. running your ls command on either the server or docker both result in āNo such file or directoryā.
However judging by the posts on this page the system that Plex is running on is completely unrelated, if your router is running miniupnpd you will run into this issue.
Sorry @ChuckPa, I thought you were specifically requesting the data from @neunon regarding his Arch Linux installation, as opposed to everyone reading the thread and experiencing the problem. You werenāt clear on that.
Iām seeing nothing matching ā*ssdp*ā in the specified folders on my XPEnology machine running Synology firmware āDSM 6.2-23739 Update 2ā.
A little further digging shows that Synology uses minissdpd which, interestingly, is coded by the same person who wrote and maintains miniupnpd. The packages are complementary and designed to operate one with the other.
I appreciate you canāt fix this Plex bug yourself, and have much bigger issues to deal with, but thanks for continuing to act as a conduit between those of us on disparate platforms (not just Arch) encountering this issue, and engineering.
As always, Iām happy to respond to any and all requests for further information and testing. Thanks for all your help thus far.
Sorry to take this posture right now but I have much bigger issues to deal with than a non-offcially supported distro.
I understand completely from a support perspective and Iād like to first of all thank you for helping us with this issue.
However, this got me thinking about whether this happened on supported distros as well, so I spun up a Fedora 29 machine I had lying around and installed a new Plex server with zero libraries and configuration whatsoever. I didnāt even claim it and make it publicly accessible. I then re-enabled UPnP on my OpenWRT router and sure enough the issue is still present even in Fedora. Plex Media Server is hitting a single core at 100%.
This issue seems to be hitting a wide variety of platforms including supported ones.
For reference, on my Fedora machine the output of ls /lib*/*ssdp* is: