Runtime instrumentation: Make runtime.uptime an Asynchronous Gauge

[cwildman]

Currently the runtime.uptime metric is reported as an Asynchronous Counter. Uptime as a concept is certainly monotonic which makes it seem like it should be a Counter, but it's also non-additive, it doesn't make sense to sum the uptime of different processes.

Additionally reporting uptime as a Counter presents a problem when trying to use delta temporality. The obvious question you want to answer with a metric like this is "How long has my process been running?". Using delta temporality with uptime would force you to track the running sum of all data points since the process started.

Because runtime.uptime is not additive and we don't actually care about rates or sum I think it is best served as an Asynchronous Gauge.

[krak3n]

In the prometheus exporter in open-telemetry/opentelemetry-collector-contrib this metric fails to get exported due to this error:

{
   "level":"error",
   "ts":1659451541.6164072,
   "caller":"prometheusexporter@v0.53.0/log.go:34",
   "msg":"error gathering metrics: collected metric runtime_uptime counter:<value:8.415e+07 >  should be a Gauge\n",
   "kind":"exporter",
   "name":"prometheus",
   "stacktrace":"github.com/open-telemetry/opentelemetry-collector-contrib/exporter/prometheusexporter.(*promLogger).Println\n\tgithub.com/open-telemetry/opentelemetry-collector-contrib/exporter/prometheusexporter@v0.53.0/log.go:34\ngithub.com/prometheus/client_golang/prometheus/promhttp.HandlerFor.func1\n\tgithub.com/prometheus/client_golang@v1.12.2/prometheus/promhttp/http.go:129\nnet/http.HandlerFunc.ServeHTTP\n\tnet/http/server.go:2084\nnet/http.(*ServeMux).ServeHTTP\n\tnet/http/server.go:2462\nnet/http.serverHandler.ServeHTTP\n\tnet/http/server.go:2916\nnet/http.(*conn).serve\n\tnet/http/server.go:1966"
}

[starzhanganz]

Can we have someone looking into this please? The incorrect type caused high error rate while exporting the metrics to Google Cloud. It reports runtime.uptime must be GAUGE, but is CUMULATIVE

[AkhigbeEromo]

Hello @MrAlias, i am currently working on a pull request for this issue. Can it be assigned to me?

[MrAlias]

This is no longer relevant after #5780. The semantic conventions for this instrumentation no longer contain this metric: open-telemetry/semantic-conventions#981.