AgentSkillsCN

database-observability

监控数据库查询、连接池，并检测N+1查询

SKILL.md
--- frontmatter
name: database-observability
description: "Instrument database queries, connection pools, and detect N+1 queries"
triggers:
  - "instrument database"
  - "trace queries"
  - "database metrics"
  - "connection pool monitoring"
  - "slow query detection"
priority: 2

Database Observability

Database is often the bottleneck. Track queries, pools, and patterns.

Query Span Attributes

AttributeExampleRequired
db.systempostgresql, mysqlYes
db.operationSELECT, INSERTYes
db.nameorders_dbYes
db.sql.tableusersRecommended
db.statementSELECT * FROM users WHERE id = ?Parameterized only!

Connection Pool Metrics (USE)

MetricTypeDescription
db.connections.activeGaugeIn use
db.connections.idleGaugeAvailable
db.connections.maxGaugePool limit
db.connections.wait_countCounterHad to wait
db.connections.wait_durationHistogramWait time

Issues to Detect

IssueDetectionFix
N+1 queries>10 identical queries per requestUse eager loading
Slow queriesDuration > p95 thresholdAdd indexes, optimize
Pool exhaustionwait_count increasingIncrease pool, fix leaks

Query Wrapper Pattern

code
Before: Start span (db.system, db.operation, db.sql.table), start timer
After:  Record duration, set db.rows_affected, record errors, end span

Pool Monitor Pattern

code
Every 10s: Record active, idle, wait_count, wait_duration from pool stats

Anti-Patterns

  • •Full SQL with values → PII risk, use parameterized only
  • •No pool metrics → Can't detect saturation
  • •Missing slow query alerts → Problems go unnoticed

References

  • •references/methodology/use-methodology.md
  • •references/platforms/{platform}/database.md