Expose read and write request latency in metrics

[lukasz-antoniak]

No description provided.

[joao-r-reis]

Some comments. This approach probably works fine but I think we should do the same thing as we are doing for the read/write metrics on the proxy metrics to be consistent.

[joao-r-reis]

the labels ...string is not very obvious, it's a list where evens are keys and odds are values for those keys... Can we make it a simple map[string]string instead? Or create a new label type:

1:

type MetricLabel struct {
   Key string
   Value string
}
func CreateHistogramNodeMetric(metricFactory MetricFactory, nodeDescription string, mn Metric, buckets []float64, labels ...MetricLabel) (Histogram, error) { }

metrics.CreateHistogramNodeMetric(metricFactory, asyncNodeDescription, metrics.AsyncRequestDuration, asyncBuckets, metrics.MetricLabel{Key: metrics.RequestDurationTypeLabel, Value: metrics.TypeReads})

2:

func CreateHistogramNodeMetric(metricFactory MetricFactory, nodeDescription string, mn Metric, buckets []float64, labels map[string]string) (Histogram, error) { }

metrics.CreateHistogramNodeMetric(metricFactory, asyncNodeDescription, metrics.AsyncRequestDuration, asyncBuckets, map[string]string{metrics.RequestDurationTypeLabel: metrics.TypeReads})

[joao-r-reis]

looks good, left just a NIT comment, feel free to ignore it if you think it's not important