kamon-prometheus
kamon-prometheus copied to clipboard
Histograms have `le` label with unformatted number
In current situation histograms are like this:
# TYPE system_metric_memory_used_bytes histogram
system_metric_memory_used_bytes{applicaton_name="token-manager-service",kamon_category="system-metric",kamon_name="memory",le="3.69098752E9"} 21.0 1497442514940
system_metric_memory_used_bytes{applicaton_name="token-manager-service",kamon_category="system-metric",kamon_name="memory",le="3.707764736E9"} 60.0 1497442514940
system_metric_memory_used_bytes{applicaton_name="token-manager-service",kamon_category="system-metric",kamon_name="memory",le="+Inf"} 60.0 1497442514940
system_metric_memory_used_bytes_count{applicaton_name="token-manager-service",kamon_category="system-metric",kamon_name="memory"} 60.0 1497442514940
system_metric_memory_used_bytes_sum{applicaton_name="token-manager-service",kamon_category="system-metric",kamon_name="memory"} 2.22113562624E11 1497442514940
With le="3.69098752E9"
, that does not make sense for labels in Prometheus.
It needs to have a configurable format, like this %6.3e
for histogram, otherwise it ends up trashing Prometheus with gazillions of buckets with unique le="#.###########E##"
.