Migrate the SLA scheduler to indexed mode

Before you begin

Back up PostgreSQL, confirm migration access, read Database migrations, and record current due transitions, missing hints, and scheduler health. Only an administrator can change mode.

Open the feature

Open Settings → Incident SLA → Support & Schedules and locate SLA Scheduler.

Configure shadow mode

1. Leave production scheduling in LEGACY while installing the optional scheduler index with the documented installer.
2. Confirm the PostgreSQL index idx_incident_next_sla_transition exists and is valid.
3. Select SHADOW. This resets clean-check counters.
4. Monitor due counts, missing scheduling hints, shadow mismatch count, errors, and job cadence.
5. Investigate every mismatch; do not treat elapsed time alone as readiness.

What OpsKnight does

LEGACY uses the legacy selection path. SHADOW compares indexed candidates with legacy truth while preserving safe behavior and populating hints. INDEXED makes the indexed path authoritative. The transition action uses an advisory lock and refuses a direct LEGACY→INDEXED jump.

Verify indexed readiness

Require the configured minimum consecutive clean checks, zero latest mismatches, a valid index, and zero active incidents missing scheduling hints. Then select INDEXED and verify due incidents transition on time without duplicate/missed SLA events.

Roll back

If indexed behavior diverges, return to SHADOW or LEGACY, preserve logs and mismatch evidence, and confirm legacy scheduling resumes. Do not drop the index during incident response; it is optional in legacy/shadow and can aid diagnosis.

Troubleshooting

• Run Shadow first: the current mode is not SHADOW.
• Resolve mismatches: latest mismatch count is nonzero.
• Wait for clean checks: the minimum consecutive comparisons has not completed.
• Install index: the required PostgreSQL index is absent/invalid.
• Missing hints: keep shadow running and inspect active incidents until the count reaches zero.

Next steps

• System logs
• Prometheus metrics