Automation Studio Health Investigation
Trace a stale, skipped, failed or degrading automation from run-level evidence into the activity that actually caused the problem.
Evidence tier describes my hands-on relationship with the content. Official sources describe platform contracts/documentation. Production observations should not automatically be treated as official limits.
Start at _AutomationInstance
Confirm the latest run status, scheduled/start/end timestamps and first error detail for the AutomationCustomerKey.
Separate Error from Skipped
Error means failure; Skipped can expose schedule overlap, queued file-drop behavior or operational timing issues.
Drill into _AutomationActivityInstance
Use AutomationInstanceID to identify the failing or long-running activity, its position and status details.
Compare runtime to baseline
A Complete run that is 2x slower than normal can be the early warning before overlap or a 30-minute Query/Script limit.
Check upstream and downstream freshness
A green automation can still produce stale/zero data. Validate critical input timestamps and output row counts.
Fix the mechanism, not the symptom
Retry once for transient contention; recurring slowness/failure should result in query design, staging, schedule or dependency changes.
Encode monitoring
Add the failure mode to the Deployment Cockpit, dashboard query, notification path or this runbook.