Pull Request Test Coverage Report for Build 947

• 85 of 87 (97.7%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.07%) to 99.082%

Totals
Change from base Build 943:	-0.07%
Covered Lines:	1726
Relevant Lines:	1742

💛 - Coveralls

Uff. Found a way to test it in a (mostly) clean way. It now fixes metrics leaking into other tests. Making metrics local wouldn't work due to name collision. Also Jaeger config is global anyway so it needs reset through protected members.

BTW would be nice to do something to avoid repeated Jaeger configuration in tests - it unnecessarily spams log output with warnings.

Nice, thanks! It was difficult to make those tests, even that way, so thanks for fixing them.

My only question is, when using the custom registry do we loose the metrics included by default?:

# HELP python_gc_objects_collected_total Objects collected during gc
# TYPE python_gc_objects_collected_total counter
python_gc_objects_collected_total{generation="0"} 2553.0
python_gc_objects_collected_total{generation="1"} 356.0
python_gc_objects_collected_total{generation="2"} 0.0
# HELP python_gc_objects_uncollectable_total Uncollectable object found during GC
# TYPE python_gc_objects_uncollectable_total counter
python_gc_objects_uncollectable_total{generation="0"} 0.0
python_gc_objects_uncollectable_total{generation="1"} 0.0
python_gc_objects_uncollectable_total{generation="2"} 0.0
...

@Agalin

PD: Btw, I'm working on migrating to opentelemetry right now. When done, will supersede this implementation, but since I don't know how long will it take, it's still good.

Those are not supported but not due to registry - instead it's an explicit limitation of the multiprocess mode.

    Registries can not be used as normal, all instantiated metrics are exported
    Custom collectors do not work (e.g. cpu and memory metrics)
    Info and Enum metrics do not work
    The pushgateway cannot be used
    Gauges cannot use the pid label

I'd assume that python gc metrics are included in memory metrics.

I see. I've never used them, and it seems more important to have the metrics be multi-process, so LGTM.