amigan
commented
6 months ago This is definitely happening right now. I have taken a heap dump hprof that I can furnish if necessary.
I have attached a heap histogram. histogram.txt
Open amigan opened 6 months ago
amigan
commented
6 months ago This is definitely happening right now. I have taken a heap dump hprof that I can furnish if necessary.
I have attached a heap histogram. histogram.txt
sdrtrunk Version master-branch
Describe the bug I noticed on a couple occasions my two tuner raspberry pi 5 8GB would spin its fan up with occasional truncated calls, and suspected GC hell. I started using a JMX prometheus exporter to take a look. I haven't gotten it reproduce that specific behavior yet, but heap usage is definitely increasing over the long term.
The middle of those graphs was a restart. I am fairly certain the blip of concurrent GC before the restart was while the system was starting to truncate calls, but I'm no longer positive. I am continuing to leave this running to collect more data, but it definitely looks like the heap is growing? I'm not a java guy, so forgive me.
I am happy to take any dumps, or enable further profiling to assist in the effort to fix this.
To Reproduce Steps to reproduce the behavior:
Expected behavior Not fans spinning up with calls truncated.
Screenshots If applicable, add screenshots to help explain your problem.
Application Log If applicable, include either the full application log,or a snippet of the error logging.
Desktop (optional - complete the following information):
Additional context Streaming to rdio-scanner running on another machine and an icecast sink. This and icecast-kh are the sole workload of this system. The P25 system is moderately busy at times but not insane. I do think the problem manifests quicker during rare traffic spikes.