B1642 potential memory leak, roon server disappearing when RAM maxed out (QNAP NAS)

Running RoonOnNas on a QNAP HS-264 /w 8GB of RAM, 55,000-70,000 tracks in library (98% local + 2% Qobuz, pretty challenging library structure). Worked stable and fairly responsive for months.

Since I updated to B1642, there seems to be some kind of runaway RAM usage by roon server process, if I keep it running for 20+hours, or do library changes (enabling/disabling storage folders, adding new local tracks).

Roon server uses like 3.6GB of RAM on start-up which is to be expected. After 24h, this was up to 7GB forcing the NAS into using virtual RAM (which I avoid as there is no SSD in either bay, roon database is on external drive). Same with roon server a few hours after starting the app, it reaches 6GB and became unresponsive/invisible for roon remote devices.

My stats are different but I have been tracking a similar pattern for many EA releases now. The result is that when log stats show memory use has increased the UX becomes poor. Zones disappear/reappear, phone app lags and hags etc.
I have to restart to get the memory usage down. UX improves immediately.

1 Like

Checked the logs, and something looks unusual to me. This pattern is repeating a zillion times, like in the region of hundreds of times per second (!) the server is on:

03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)
03/06 16:36:45 Trace: [radio/query] performing channel query
03/06 16:36:45 Trace: [radio/query] query returned 18 Sooloos.Broker.Api.Channel(s)

Any idea what is going on here?

Same here, but depending on what I do with the roon server, it works only for an hour or a few hours until the memory leak hits the ceiling again.

Interesting. Aren’t Production and EA builds the same after the latest update?

How many of these are there with the same timestamp?

Do you have Roon radio on and using it?

These are fairly benign traces, but the quantity could be Roon Radio is constantly adding tracks to the queue, even if not shown.

If this theory is correct then it’ll consume memory.

I haven’t seen this the last time I ran EA or this time around on the production build.

1 Like

1,300 per second, but it might repeat every second for quite some time.

It is active, but I do not use it very often, just in case an album or playlist runs out of tracks.

Thanks for your hint, will deactivate roon radio, clear all queues and restart roon server application.

Deactivating roon radio and clearing the queues changed nothing.

But by chance I might have found a solution: Noticed that the roon server application was constantly busy, eating between 25% and 50% of CPU power constantly, with all background tasks deactivated. Normally it stays well below 0.5% with this machine in idle.

Tried to narrow down potential causes, and removed all devices from the network, one by one. When removing a third computer (MacBook Pro) running roon as remote but never really managing to connect to roon server, all issues stopped immediately, CPU usage is down to an average 0.4%.

To me it looks like roon (as remote) on this computer permanently tried to connect to the roon server but failed (for whatever reason) and keeping it busy. Let us see if this was the cause of the memory leak as well.

My logs are showing aggressive GC.

12gb to 8gb

This was with a 400k track library.

Maybe you have highlighted a potential cause.

@ncpl , how does this stack up at your end?

Update: All stable and fine now with RAM consumption at 3.7GB and CPU usage at 0.4% while streaming, after removing one of the two computers running roon as remote from the network.

It seems that roon server was getting confused trying to get in touch with both remote devices permanently at the same time.