hard20 pts · first answer countsprometheuslatencyslis
Misleading histogram quantiles
Your latency SLO is "99% of requests under 300 ms". The histogram buckets are 0.1, 0.5, 1 and 5 seconds. Why is an alert on histogram_quantile(0.99, ...) > 0.3 unreliable?
Discussion 0
Join free to post in the discussion.
Related questions
- How Prometheus collects metricsSite Reliability Engineer
- Writing an SLI as a ratioSite Reliability Engineer
- Counter or gauge?Site Reliability Engineer
- Using rate() on a counterSite Reliability Engineer
- Percentiles vs average latencySite Reliability Engineer
- p99 latency across all instancesSite Reliability Engineer
- Alerting when a job disappearsSite Reliability Engineer
- The for clause in alert rulesSite Reliability Engineer