CPU stuck at ~70%

[Callumpy]

Sonarr version (exact version): 2.0.0.5338
Mono version (if Sonarr is not running on Windows): 5.20.1.34
OS: unRAID 6.6.6
Trace logs: https://drive.google.com/file/d/1nknXTG9l063JzfAlOhTvmIzJB3AADG1C/view?usp=sharing
Description of issue:

I have Sonarr and Radarr installed in UnRAID using linuxserver’s Docker containers, and they work great most of the time.

When I leave them running, I tend to find either in the morning or after I get home from work, the CPU has been pegged up over 70% for either container (sometimes both). Something seems to be causing the CPU to ramp up, and then it never dies.

Restarting the container returns them back to normal usage (~1%).

Thanks for reading!

[Callumpy]

It’s happened again overnight, more logs:

https://drive.google.com/file/d/1lrexKvSYoxnJHQyqkPVADMAG9XiwbFWa/view?usp=sharing

[Callumpy]

Still ongoing, here’s some more details from Radarr this time.

Here are some graphs from my netdata install, showing that at ~3:39 in the morning, the CPU ramped up and stayed there until I restarted it about 5 minutes ago:

Here are the debug and trace logs for that time: https://drive.google.com/drive/folders/1s5mBaNre2N7xky26bBNuIgIj9CO5XHWT?usp=sharing

[markus101]

Let’s focus on the Sonarr side of things.

The first thing would be trace logs covering the initial spike of CPU along with the corresponding graph, without knowing when it spiked the trace logs haven’t been too helpful yet.

[Taloth]

Radarr restarted:

20-2-21 02:59:32.9|Trace|CommandExecutor|CheckForFinishedDownloadCommand <- DownloadMonitoringService [00:00:00.0351770]
20-2-21 03:00:02.8|Trace|Scheduler|Pending Tasks: 0
20-2-21 03:01:23.5|Debug|ProcessProvider|Found 0 processes with the name: Radarr.Console
20-2-21 03:01:23.5|Debug|ProcessProvider|Found 1 processes with the name: Radarr

Around 3:50 all indexer requests failed with time outs. Could be related with the real issue, but nothing weird after that and queries to sabnzbd went fine. I wonder if it’s some previously undiscovered mono issue in mono 5.20.
Either way, correlated that with logs of other incidents.

[Callumpy]

Thanks for your response!

That’s very fair, I’ll try to catch it again now I’ve got netdata running.

[Callumpy]

Thanks for your response too!

If there’s anything you can think of that I can take a look at/grab logging for, I’d be happy to do so if it helps identify an issue!

I forgot to mention that during the CPU peaks, a htop run will show that mono with either nzbdrone.exe or radarr.exe are the culprits.

[Taloth]

The ball is in your court atm. We already mentioned you need to collect sonarr logs and look at the time that the cpu usage starts.
Check that vs the radarr logs around similar cpu peaks. There’s a commonality that you need to identify.