在 Keep 中,告警被视为一等公民,具有明确定义的严重程度和状态,以帮助快速高效地响应。
告警严重程度
Keep 中的告警严重程度分为五类,帮助团队根据告警的紧急程度和影响优先处理响应。
| 严重程度级别 | 描述 | 期望值 |
|---|---|---|
| CRITICAL | 需要立即行动。 | "critical" |
| HIGH | 需要尽快处理。 | "high" |
| WARNING | 表示潜在问题。 | "warning" |
| INFO | 提供信息,无需立即行动。 | "info" |
| LOW | 次要问题或最低优先级。 | "low" |
告警状态
Keep 中告警的状态反映了其在告警生命周期中的当前状态。
| 状态 | 描述 | 期望值 |
|---|---|---|
| FIRING | 表示持续问题的活动告警。 | "firing" |
| RESOLVED | 问题已解决,告警不再活动。 | "resolved" |
| ACKNOWLEDGED | 告警已被确认但未解决。 | "acknowledged" |
| SUPPRESSED | 告警因各种原因被抑制。 | "suppressed" |
| PENDING | 数据不足或数据不足以确定告警状态。 | "pending" |
提供者告警映射
不同的提供者可能有其特定的定义和处理告警严重程度和状态的方式。 Keep 通过将它们映射到定义的枚举(AlertSeverity 和 AlertStatus)来标准化这些变体。
以下是各种提供者如何与 Keep 的告警系统对齐:
| 提供者 | 严重程度映射 | 状态映射 |
|---|---|---|
| CloudWatch | N/A | ALARM -> FIRING, OK -> RESOLVED, INSUFFICIENT_DATA -> PENDING |
| Prometheus | "critical" -> CRITICAL "warning" -> WARNING, "info" -> INFO, "low" -> LOW | "firing" -> FIRING, "resolved" -> RESOLVED |
| Datadog | "P4" -> INFO, "P3" -> WARNING, "P2" -> HIGH, "P1" -> CRITICAL | "Triggered" -> FIRING, "Recovered" -> RESOLVED, "Muted" -> SUPPRESSED |
| PagerDuty | "P1" -> CRITICAL, "P2" -> HIGH, "P3" -> WARNING, "P4" -> INFO | "triggered" -> FIRING, "acknowledged" -> ACKNOWLEDGED, "resolved" -> RESOLVED |
| Pingdom | N/A | "down" -> FIRING, "up" -> RESOLVED, "paused" -> SUPPRESSED |
| Dynatrace | "critical" -> CRITICAL, "warning" -> WARNING, "info" -> INFO | "open" -> FIRING, "closed" -> RESOLVED, "acknowledged" -> ACKNOWLEDGED |
| Grafana | "critical" -> CRITICAL, "high" -> HIGH, "warning" -> WARNING, "info" -> INFO | "ok" -> RESOLVED, "paused" -> SUPPRESSED, "alerting" -> FIRING, "pending" -> PENDING, "no_data" -> PENDING |
| New Relic | "critical" -> CRITICAL, "warning" -> WARNING, "info" -> INFO | "open" -> FIRING, "closed" -> RESOLVED, "acknowledged" -> ACKNOWLEDGED |
| Sentry | "fatal" -> CRITICAL, "error" -> HIGH, "warning" -> WARNING, "info" -> INFO, "debug" -> LOW | "resolved" -> RESOLVED, "unresolved" -> FIRING, "ignored" -> SUPPRESSED |
| Zabbix | "not_classified" -> LOW, "information" -> INFO, "warning" -> WARNING, "average" -> WARNING, "high" -> HIGH, "disaster" -> CRITICAL | "problem" -> FIRING, "ok" -> RESOLVED, "acknowledged" -> ACKNOWLEDGED, "suppressed" -> SUPPRESSED |