Memory leak in mono-sgen process?

I have a raspberry pi with a hat running the roonbridge software. Every couple of months I log into my pi and the mono-sgen process has eaten all the available ram on the device. Killing the process frees it back up, but then it just happens again.

I’m not even playing anything with this bridge, it just sits there as I use it very occasionally. However, it’s always needing to be killed to free device memory.

Anyone else see this?

Tuning garbage collectors like mono-sgen is more of an art than a science, I’m afraid. What OS are you running on your Pi?

DietPi v8.3.1 on a RPi 3 Model B (aarch64)

Nobody else has this problem? Is the only solution really to just kill -9 it every few weeks?

Maybe try a newer release?

Shouldn’t having the latest roon bridge client update that too? If it’s a dependent package, it should have the updates, right?

I meant of diet pi. I think yours is a year old.

I updated to 8.19.1, hopefully it’ll perform better. Thanks.

Nope, it’s still happening.

I think it is a long time since Roon Bridge was updated. The current version is still 1.8 build 1125 which dates back to September 2022

This version of DietPi is somewhat old (April 2022).

This version is still quite old.

The current version is 9.8.0.

It is very easy to update Dietpi from a terminal using ‘dietpi-update’.

I’ve not had the best luck updating in large steps with dietpi-update, it’ll often fail and leave the device in a weird state. This is especially true of my USBridge.

Well certainly, the USBridge is outside of my experience. For USB connected DACs, I just use the USB ports built onto the Raspberry Pi 4 itself. As a consequence, I can’t comment on any issues related to the USBridge that may be encountered if you did upgrade.

The only time that I had an issue with dietpi-update was when there was a change to the boot folder structure which mean’t that the rpi-eeprom package was pulling the wrong version of one of its depdencies which occured in version 8.23 (I think). This was resolved in version 8.24 and I have had no issues since. Note: I am, and have always been, on the Bookworm kernel.