mirror of
https://github.com/makayabou/asg-server.git
synced 2026-05-02 17:43:36 +02:00
57 lines
2.5 KiB
YAML
57 lines
2.5 KiB
YAML
groups:
|
|
- name: http-alerts
|
|
rules:
|
|
- alert: HighHTTPErrorRateWarning
|
|
expr: sum(rate(http_requests_total{job="backend", status_code=~"5.."}[5m])) / sum(rate(http_requests_total{job="backend"}[5m])) > 0.05
|
|
for: 5m
|
|
labels:
|
|
severity: warning
|
|
annotations:
|
|
summary: "High HTTP error rate (Warning)"
|
|
description: "The HTTP error rate has exceeded 5% over the last 5 minutes."
|
|
|
|
- alert: HighHTTPErrorRateCritical
|
|
expr: sum(rate(http_requests_total{job="backend", status_code=~"5.."}[5m])) / sum(rate(http_requests_total{job="backend"}[5m])) > 0.1
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: "High HTTP error rate (Critical)"
|
|
description: "The HTTP error rate has exceeded 10% over the last 5 minutes."
|
|
|
|
- alert: HighHTTPLatencyWarning
|
|
expr: histogram_quantile(0.99, sum(rate(http_request_duration_seconds_bucket{job="backend"}[5m])) by (le)) > 0.3
|
|
for: 5m
|
|
labels:
|
|
severity: warning
|
|
annotations:
|
|
summary: "High HTTP latency (p99) (Warning)"
|
|
description: "The p99 HTTP latency has exceeded 0.3 seconds over the last 5 minutes."
|
|
|
|
- alert: HighHTTPLatencyCritical
|
|
expr: histogram_quantile(0.99, sum(rate(http_request_duration_seconds_bucket{job="backend"}[5m])) by (le)) > 1
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: "High HTTP latency (p99) (Critical)"
|
|
description: "The p99 HTTP latency has exceeded 1 seconds over the last 5 minutes."
|
|
|
|
- alert: IncreasedHTTPRequestVolumeWarning
|
|
expr: rate(http_requests_total{job="backend"}[1m]) > 2 * avg_over_time(rate(http_requests_total{job="backend"}[1m])[10m:1m])
|
|
for: 5m
|
|
labels:
|
|
severity: warning
|
|
annotations:
|
|
summary: "Increased HTTP request volume (Warning)"
|
|
description: "The HTTP request volume has increased by 2x compared to the average of the last 10 minutes."
|
|
|
|
- alert: IncreasedHTTPRequestVolumeCritical
|
|
expr: rate(http_requests_total{job="backend"}[1m]) > 5 * avg_over_time(rate(http_requests_total{job="backend"}[1m])[10m])
|
|
for: 5m
|
|
labels:
|
|
severity: critical
|
|
annotations:
|
|
summary: "Increased HTTP request volume (Critical)"
|
|
description: "The HTTP request volume has increased by 5x compared to the average of the last 10 minutes."
|
