Latency SLOs - Duplicated "p100" percentile in duration graph (UI)

[svenmueller]

The duration graph for latency SLOs shows a duplicated "p100" percentile label.

Version 0.7.6

Configuration

apiVersion: pyrra.dev/v1alpha1
kind: ServiceLevelObjective
metadata:
  labels:
    prometheus: k8s
    role: alert-rules
  name: xyz
  namespace: monitoring
spec:
  alerting: {}
  description: 99.9% of xyz requests in the past 28 days are faster than 100
    ms.
  indicator:
    latency:
      grouping:
      - destination_service_name
      success:
        metric: istio_request_duration_milliseconds_bucket{job="kubernetes-pods",
          le="100"}
      total:
        metric: istio_request_duration_milliseconds_count{job="kubernetes-pods"}
  target: "99.9"
  window: 28d
status:
  type: PrometheusRule

[metalmatze]

Interesting find! It seems like there are some floating point rounding errors in place which cause both 0.9990000000000001 and 0.999 to be added to the list of percentiles.

[metalmatze]

This is now fixed and https://github.com/pyrra-dev/pyrra/releases/tag/v0.7.7 released with it.