fix: guard configurable monitoring ratios

This commit is contained in:
hectorzhao
2026-08-14 17:59:22 +08:00
parent 6ccc102830
commit 0cd09441da
4 changed files with 12 additions and 7 deletions
+4 -4
View File
@@ -32,12 +32,12 @@ groups:
labels: { severity: critical, service: host }
annotations: { summary: "根磁盘使用率达到严重阈值", description: "根磁盘使用率持续超过90%。", currentValue: "{{ $value }}", threshold: "90%" }
- alert: CmppApiHttpErrorRateWarning
expr: (100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) > 1) and (100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) <= 5)
expr: (100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) > 1) and (100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) <= 5) and (sum(increase(cmpp_api_http_requests_total{status=~"5.."}[5m])) >= 5)
for: 5m
labels: { severity: warning, service: api }
annotations: { summary: "API 5xx 错误率达到警告阈值", description: "API 5xx 错误率持续超过1%。", currentValue: "{{ $value }}", threshold: "1%" }
- alert: CmppApiHttpErrorRateCritical
expr: 100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) > 5
expr: (100 * sum(rate(cmpp_api_http_requests_total{status=~"5.."}[5m])) / clamp_min(sum(rate(cmpp_api_http_requests_total[5m])), 0.001) > 5) and (sum(increase(cmpp_api_http_requests_total{status=~"5.."}[5m])) >= 5)
for: 5m
labels: { severity: critical, service: api }
annotations: { summary: "API 5xx 错误率达到严重阈值", description: "API 5xx 错误率持续超过5%。", currentValue: "{{ $value }}", threshold: "5%" }
@@ -82,12 +82,12 @@ groups:
labels: { severity: critical, service: postgresql }
annotations: { summary: "PostgreSQL 连接使用率达到严重阈值", description: "PostgreSQL 连接使用率持续超过85%。", currentValue: "{{ $value }}", threshold: "85%" }
- alert: RedisMemoryWarning
expr: (100 * redis_memory_used_bytes / redis_memory_max_bytes > 70) and (100 * redis_memory_used_bytes / redis_memory_max_bytes <= 85)
expr: (100 * redis_memory_used_bytes / redis_memory_max_bytes > 70) and (100 * redis_memory_used_bytes / redis_memory_max_bytes <= 85) and (redis_memory_max_bytes > 0)
for: 10m
labels: { severity: warning, service: redis }
annotations: { summary: "Redis 内存使用率达到警告阈值", description: "Redis 内存使用率持续超过70%。", currentValue: "{{ $value }}", threshold: "70%" }
- alert: RedisMemoryCritical
expr: 100 * redis_memory_used_bytes / redis_memory_max_bytes > 85
expr: (100 * redis_memory_used_bytes / redis_memory_max_bytes > 85) and (redis_memory_max_bytes > 0)
for: 5m
labels: { severity: critical, service: redis }
annotations: { summary: "Redis 内存使用率达到严重阈值", description: "Redis 内存使用率持续超过85%。", currentValue: "{{ $value }}", threshold: "85%" }