Prometheus receiver prints "level" field twice in log messages

[nilebox]

Describe the bug
All log messages produced by Prometheus code imported by Prometheus receiver print "level" field twice, e.g.:

{"level":"info","ts":1596545757.1352987,"caller":"scrape/scrape.go:962","msg":"","level":"debug","scrape_pool":"target2","target":"http://127.0.0.1:38363/target2/metrics","msg":"Scrape failed","err":"server returned HTTP status 404 Not Found"}

The problem is in the zap -> go-kit log adapter:

opentelemetry-collector/receiver/prometheusreceiver/internal/logger.go

Line 37 in bd921af

w.l.Infow("", keyvals...)

• l.Infow() produces the first "level":"info" field
• the "level":"debug" in the example above is produced by go-kit logger (https://github.com/go-kit/kit/blob/556100560949062d23fe05d9fda4ce173c30c59f/log/level/level.go#L178-L188) and passed as part of "keyvals" to the adapter linked above.

Steps to reproduce

1. Replace logger with zap.NewProduction() in

   opentelemetry-collector/receiver/prometheusreceiver/metrics_receiver_test.go

   Line 45 in bd921af

   var logger = zap.NewNop()

2. Run EndToEndTest:

   opentelemetry-collector/receiver/prometheusreceiver/metrics_receiver_test.go

   Lines 928 to 929 in bd921af

   	// TestEndToEnd end to end test executor
   	func TestEndToEnd(t *testing.T) {

What did you expect to see?
A single "level" field (probably the one produced by Prometheus code / go-kit)
To implement this fix we need to improve the function

opentelemetry-collector/receiver/prometheusreceiver/internal/logger.go

Line 34 in bd921af

func (w *zapToGokitLogAdapter) Log(keyvals ...interface{}) error {

, which should check if level is one of the fields in keyvals and respect it for using appropriate zap level, and then drop the go-kit one.

What did you see instead?
Two "level" fields in JSON output

[nilebox]

There are other duplicate fields in the log:

• "msg":"" and "msg":"Scrape failed"
• "ts":1596117234.8600261 and "timestamp":1596117234850