Does upnp create logging threads? 10 million junp and 20 million logger threads

Setup, Configuration and Use Runtime
1

On the latest 5.0.2.-1, after a few days of running. I have sonos and wemo. Jvm itself over 4Gb after a few days…. I know about the upnp issues but was surprised to see all these logging threads.

num instances #bytes class name (module)

1: 20725134 497403216 org.ops4j.pax.logging.slf4j.Slf4jLogger
2: 10360545 331537440 org.jupnp.protocol.async.SendingSearch
3: 10360537 331537184 java.util.concurrent.LinkedTransferQueue$DualNode (java.base@21.0.8)
4: 10360575 165769200 org.jupnp.model.types.UDN
5: 10360545 165768720 org.jupnp.model.message.header.UDNHeader

2

Hello @billfor,

Does your log4j configuration is the default one, or did you increase log level for upnp ?
I don’t remember to see such an issue before.

Laurent.

3

It’s not default but nothing that I changed would have anything to do with the high thread count. There was no activity in the logs during the time when I took the histogram. I don’t do anything special for upnp.

I know about the upnp issues but was just surprised to see that log4j was somehow affected. 20 million is just huge. It was using over 4Gb on my system after forced GC. I’ve disabled all the upnp stuff for now and everything works fine, except for my sonos and wemo stuff.

4

I will try to check if I see something comparable on my side.
But since the last fixes, and the recent snapshot beginning early in November, I’ve don’t experience anymore issue with upnp. My openhab installation is stable for at least 7-10 days, memory not growing to much, and everything works ok : Sonos, etc.

I ask this because I was thinking that upnp regression was introduce after 5.0.2.
So this is perhaps another issue, but not the one fixes recently.

5

I see the M‑SEARCH changes I didn’t realize they were supposed to be in this version. I do also have a WiiM device which has always has issues with the binding. Fortunately WiiM has an http API that can be used instead.

I will re-enable it after a few more days I just want to see if everything is stable with it disabled. What I didn’t understand was why the 20million Slf4jLogger threads should be associated with the upnp issue.