SQL Server 检查
sqlserver 模块有 13 项检查。VM 或本地部署上的 SQL Server:实例的运行状况、性能和容量(T-SQL,使用具有 VIEW SERVER STATE 权限的登录名)。
本页在构建文档时根据产品的检查目录生成,因此规则、间隔和默认严重性始终与你正在阅读的版本一致。阈值是默认值;实体和资源可以覆盖这些阈值。
| 检查 | 间隔 | 默认严重性 | 规则 |
|---|---|---|---|
sqlserver.cpu_percentCPU utilization (SQL process) | 60s | warning | warning > 80 percent · 15m / critical > 90 percent · 5m |
sqlserver.ple_secondsPage life expectancy | 60s | warning | warning < 1800 seconds · 10m / critical < 600 seconds · 5m |
sqlserver.batch_requestsBatch requests per second | 60s | info | ±2σ over 14d (learning only) |
sqlserver.blocked_sessionsBlocked sessions | 30s | warning | warning >= 1 count · 30s / critical >= 5 count · 60s |
sqlserver.deadlocks_per_minDeadlocks per minute | 60s | warning | warning > 0 per_minute / critical >= 5 per_minute |
sqlserver.top_waitsTop wait categories | 60s | info | ±2σ over 14d (learning only) |
sqlserver.log_space_percentTransaction log space used | 300s | warning | warning > 75 percent / critical > 90 percent |
sqlserver.backup_age_fullLast full backup age | 600s | warning | warning > 1560 minutes / critical > 2880 minutes |
sqlserver.backup_age_logLast log backup age | 300s | warning | warning > 30 minutes / critical > 60 minutes |
sqlserver.agent_jobs_failedSQL Agent job failures (last 24 h) | 300s | warning | warning >= 1 count / critical >= 3 count |
sqlserver.errorlog_severityError log entries with severity 16+ | 120s | warning | warning >= 1 count / critical >= 5 count |
sqlserver.build_versionBuild and update level | 3600s | info | any change |
sqlserver.linked_server_pingLinked server reachability and latency | 300s | critical | warning > 1000 milliseconds / critical > 5000 milliseconds · 15m |
检查
sqlserver.cpu_percent
CPU utilization (SQL process)
- 度量
cpu_percent(percent) · 每60s·tsql- 规则
warning > 80 percent · 15m / critical > 90 percent · 5m· 默认严重性: warning- 措施
- —
- 基线
- Sustained CPU above 80% warrants investigation; above 90% for minutes is critical on production. · 来源
sqlserver.ple_seconds
Page life expectancy
- 度量
ple_seconds(seconds) · 每60s·tsql- 规则
warning < 1800 seconds · 10m / critical < 600 seconds · 5m· 默认严重性: warning- 措施
- —
- 基线
- Rule of thumb: 300 s per 4 GB of buffer pool; defaults assume a 24 GB pool. Adjust per instance.
sqlserver.batch_requests
Batch requests per second
- 度量
batch_requests_per_sec(per_second) · 每60s·tsql- 规则
±2σ over 14d (learning only)· 默认严重性: info- 措施
- —
- 基线
- No fixed threshold; adaptive baseline flags unusual throughput for this instance.
sqlserver.blocked_sessions
Blocked sessions
- 度量
blocked_sessions(count) · 每30s·tsql- 规则
warning >= 1 count · 30s / critical >= 5 count · 60s· 默认严重性: warning- 措施
- Identify the head blocker (sys.dm_exec_requests / sp_WhoIsActive) before killing anything.
- 基线
- Any blocking lasting more than 30 s deserves attention; chains of 5+ sessions are critical.
sqlserver.deadlocks_per_min
Deadlocks per minute
- 度量
deadlocks_per_min(per_minute) · 每60s·tsql- 规则
warning > 0 per_minute / critical >= 5 per_minute· 默认严重性: warning- 措施
- —
- 基线
- Azure Monitor recommended rule for SQL: alert on any deadlock.
sqlserver.top_waits
Top wait categories
- 度量
wait_time_ms(milliseconds) · 每60s·tsql- 规则
±2σ over 14d (learning only)· 默认严重性: info- 措施
- —
- 基线
- Delta-based; the timeline drill-down uses this series.
sqlserver.log_space_percent
Transaction log space used
- 度量
log_space_percent(percent) · 每300s·tsql- 规则
warning > 75 percent / critical > 90 percent· 默认严重性: warning- 措施
- —
- 基线
- Log at 75% means the next growth or a long transaction may fill the volume.
sqlserver.backup_age_full
Last full backup age
- 度量
full_age_minutes(minutes) · 每600s·tsql- 规则
warning > 1560 minutes / critical > 2880 minutes· 默认严重性: warning- 措施
- —
- 基线
- Daily full: warn at 26 h, critical at 48 h. Tie to the entity's RPO policy. A database never backed up counts as infinitely old.
sqlserver.backup_age_log
Last log backup age
- 度量
log_age_minutes(minutes) · 每300s·tsql- 规则
warning > 30 minutes / critical > 60 minutes· 默认严重性: warning- 措施
- —
- 基线
- RPO 15 min → warn at 30, critical at 60.
sqlserver.agent_jobs_failed
SQL Agent job failures (last 24 h)
- 度量
failed_jobs(count) · 每300s·tsql- 规则
warning >= 1 count / critical >= 3 count· 默认严重性: warning- 措施
- —
- 基线
- Any failed job is at least a warning; backup/maintenance jobs can be promoted to critical by tag.
sqlserver.errorlog_severity
Error log entries with severity 16+
- 度量
errorlog_entries(count) · 每120s·tsql- 规则
warning >= 1 count / critical >= 5 count· 默认严重性: warning- 措施
- —
- 基线
- Severity 16+ indicates errors the application must handle; 20+ are fatal. Counts entries of the last 5 minutes.
sqlserver.build_version
Build and update level
- 度量
build_info· 每3600s·tsql- 规则
any change· 默认严重性: info- 措施
- —
- 基线
- Inventory only; the updates module compares it with the latest CU/GDR.
sqlserver.linked_server_ping
Linked server reachability and latency
- 度量
latency_ms(milliseconds) · 每300s·tsql- 规则
warning > 1000 milliseconds / critical > 5000 milliseconds · 15m· 默认严重性: critical- 措施
- Check network path (VPN tunnel, firewall) first; then the remote service; then provider credentials on the linked server.
- 基线
- Warn when a remote round trip exceeds 1 s; 3 consecutive failures or timeouts are critical.
相关内容
- SQL Server:先决条件以及如何添加资源
- 阈值和基线:如何覆盖默认阈值