# Known Incident Patterns

TODO: Document P4 server incident signatures as they are identified and validated.

## Pattern: Database Wedge

**Description:** Write lock on a db.* table (commonly db.rev, db.have, db.working)
causes cascading pile-up of read-waiting processes.

**Log signature:**
- One process with high `write held` time on a table
- Multiple processes with high `read wait` time on the same table
- `p4 monitor show -ale` shows many W (waiting) processes

**Detection:** `WedgeDetector` in `detector.py`

**Corrective actions (in tier order):**
- Tier 1: Alert on-call with diagnosis and table name
- Tier 4 (if pre-authorized): `p4 monitor terminate <pid>` on lock-holder

## Pattern: Slow Submit

**Description:** Single `p4 submit` with long compute/lock time, no pile-up.

**Detection:** `SlowCommandDetector`

## Pattern: Checkpoint Under Load

**Description:** `p4 admin checkpoint` holding broad locks while users are active.

**Detection:** `WedgeDetector` (checkpoint holds write locks on multiple tables)