sql-server-investigation

name	sql-server-investigation
description	Incident investigation procedures for Amazon RDS for SQL Server performance issues including blocking and lock contention, plan regression, slow or high-impact queries, and missing indexes. Use this skill when investigating SQL Server database latency, high CPU, blocked sessions, query timeouts, or read/write performance degradation. Triggers on requests like "investigate SQL Server blocking", "RDS SQL Server high CPU", "query timeout RCA", "plan regression", or "why is my SQL Server slow".
metadata	{"version":"1.0.0","aws-devops-agent-skills.agent-types":"Incident RCA","aws-devops-agent-skills.aws-services":"Amazon RDS","aws-devops-agent-skills.technical-domains":"Databases"}

SQL Server Investigation

Structured troubleshooting methodology for Amazon RDS for SQL Server. Work from the lightweight signal outward: use native AWS APIs to triage and localize the problem, then use the MCP tools to resolve the session-, plan-, and index-level detail those APIs cannot reach.

Investigation Workflow

Follow these steps in order for every investigation.

Step 1: TRIAGE — Assess Overall Health

Always start here regardless of the reported symptom. Retrieve current values and active alarms from native AWS APIs:

cloudwatch.DescribeAlarms     # active alarms on the instance
cloudwatch.GetMetricData      # CPUUtilization, DatabaseConnections,
                              # FreeableMemory, FreeStorageSpace
pi.GetResourceMetrics         # db.load — Average Active Sessions (AAS)

Severity thresholds:

Severity	Criteria
CRITICAL	CPU > 90% OR AAS > 8 OR connections near max OR FreeableMemory < 1 GB