3aca674a78
앞선 감사 네 번은 "이 코드가 제대로 도는가"를 물었다. 이번엔 작업자·노조·안전관리자·
공장장·설치기사 5인 관점으로 3단논법을 세우고 전제를 코드로 확인한 뒤 반문으로 깼다.
5인 중 2인의 논증이 완전히, 2인이 부분적으로 무너졌다.
읽기 감사 — 이 시스템은 쓰기를 전부 감사했다. 임계값 한 번 바꾼 것까지 해시 체인에
들어간다. 그런데 작업자 사진을 열어보는 것은 하나도 남지 않았다. 감시 시스템과 안전
시스템을 가르는 선이 정확히 거기라서, 나머지 프라이버시 설계(RAM 링버퍼 7.5초·전송 전
얼굴 블러·라이브뷰 부재)가 전부 이 한 줄이 없어서 검증 불가능한 주장에 머물러 있었다.
거부된 열람은 기록하지 않는다 — 그러면 감사 로그가 "본 사람"이 아니라 "요청한 사람"
목록이 되어 정작 누가 봤는지를 묻는 데 못 쓴다.
크론 생존 — 죽은 노드는 스스로 신고할 수 없어서 크론이 하트비트의 부재로 판정한다.
그런데 그 크론이 멈추면 아무도 판정하지 않는다. 통보·5분 확대·장치 이상 발행이 전부 그
한 프로세스에 걸려 있어서, 멈춘 순간부터 화면은 조용해지고 그 침묵은 "사고 없음"과
구별되지 않는다. 감시자의 생존은 감시자가 아닌 쪽이 읽어야 한다.
증거 보존 — 스냅샷 삭제가 나이 하나만 봤다. 기본 7일인데 중대재해 조사는 몇 주에서 몇
달이라, 우리가 내건 목적("사고 후 원인 분석")이 8일째에 스스로 사라졌다. 확인 처리로
자동 연장하지 않는 이유는 자동이면 사실상 전량 보존이 되어 7일 약속이 무너지기 때문이다.
증거 정책 — 프레임이 ergo_report 하나만 빼고 모든 이벤트에 붙었다. 체류 상한 초과도
포함이라, 밀폐공간 프리셋 기본값(30분) 기준으로 정상 작업자가 규정대로 일하는 동안
주기적으로 본인 사진이 서버로 갔다. 위험해서가 아니라 오래 있어서. 기준을 명문화했다:
시간 경과 자체가 조건인 이벤트에는 사진을 붙이지 않는다.
마이그레이션 2026-09-25-evidence-hold-and-view-audit.sql (README-UPLOAD 10번).
검증: phpunit 302 passed(신규 18, 스킵 0) · pytest(신규 18)
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01EBjuR7DW6PNNMows4KnbYU
460 lines
27 KiB
SQL
460 lines
27 KiB
SQL
-- cloud/schema.sql
|
|
-- MySQL 8.0+ DDL for the P1-B cloud ingest + dashboard subset
|
|
-- (backing store for App\Repository\Pdo\* -- see cloud/src/Repository/Pdo/).
|
|
--
|
|
-- Scope: this is the P1-B subset of the full design-spec schema (design doc §7.4).
|
|
-- Tables/columns that belong to later phases are intentionally OMITTED here (YAGNI):
|
|
-- sites, escalations, oncall, compliance, retention_policy, audit hash-chaining
|
|
-- (prev_hash/hash on audit_log), and events.disposition/is_test/legal_hold/lifecycle_state.
|
|
-- See docs/superpowers/specs/2026-07-11-workplace-safety-ai-cctv-design.md §7.4 and the
|
|
-- Task 16 brief self-review for the P1-B / P2 / P3 split.
|
|
--
|
|
-- sqlite compatibility notes (cloud/tests/PdoIntegrationTest.php uses its own minimal
|
|
-- sqlite-flavored DDL for the tables it round-trips -- this file targets MySQL only,
|
|
-- these notes exist to explain the mechanical translation if that ever needs to grow):
|
|
-- * `INT ... AUTO_INCREMENT` -> sqlite: `INTEGER PRIMARY KEY AUTOINCREMENT` (must be the
|
|
-- table's sole INTEGER PRIMARY KEY column)
|
|
-- * `ENGINE=InnoDB` -> sqlite: omit (no storage engines)
|
|
-- * `DEFAULT CHARSET=...` -> sqlite: omit
|
|
-- * `JSON` column type -> sqlite: `TEXT` (no native JSON type; app always
|
|
-- reads/writes JSON-encoded strings either way)
|
|
-- * `DEFAULT (UNIX_TIMESTAMP())` -> sqlite: `DEFAULT (strftime('%s','now'))`
|
|
-- * `INSERT ... ON DUPLICATE KEY UPDATE` (rate_limits upsert) -> sqlite:
|
|
-- `INSERT ... ON CONFLICT(rate_key, window_start)
|
|
-- DO UPDATE SET count = count + 1`
|
|
-- * `CHECK (...)` on enum-like columns -> sqlite supports CHECK natively, no change
|
|
-- * `TINYINT(1)` -> sqlite: `INTEGER` (0/1)
|
|
-- * `BIGINT UNSIGNED` -> sqlite: `INTEGER` (sqlite ints are always 64-bit signed)
|
|
--
|
|
-- All app-level SQL against this schema uses prepared statements with `?` placeholders --
|
|
-- no string interpolation of values anywhere in App\Repository\Pdo\*.
|
|
|
|
SET NAMES utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- users: dashboard accounts (auth + RBAC + brute-force lockout).
|
|
-- No PdoUserRepository::create() exists in P1-B -- the first admin account is
|
|
-- provisioned out-of-band via the CLI snippet in cloud/docs/cloud-deploy.md.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_users (
|
|
id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
username VARCHAR(190) NOT NULL,
|
|
password_hash VARCHAR(255) NOT NULL, -- Argon2id, see App\Security\Password::hash()
|
|
role VARCHAR(32) NOT NULL, -- app-whitelisted: admin|operator|viewer (App\Auth\Rbac)
|
|
site_scope VARCHAR(190) NULL,
|
|
failed_attempts INT UNSIGNED NOT NULL DEFAULT 0,
|
|
locked_until BIGINT UNSIGNED NOT NULL DEFAULT 0, -- unix timestamp; 0 = not locked
|
|
created_at BIGINT UNSIGNED NOT NULL DEFAULT (UNIX_TIMESTAMP()),
|
|
PRIMARY KEY (id),
|
|
UNIQUE KEY uq_users_username (username),
|
|
CONSTRAINT chk_users_role CHECK (role IN ('admin','operator','viewer'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- devices: one row per Pi/edge device; holds the per-device HMAC ingest secret.
|
|
-- No PdoDeviceRepository::create() exists -- rows are provisioned by an admin
|
|
-- (per-device secret issuance) before the device is allowed to sync/ingest.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_devices (
|
|
device_id VARCHAR(64) NOT NULL,
|
|
name VARCHAR(190) NULL,
|
|
site VARCHAR(64) NULL, -- access-control scope; matches users.site_scope (NULL = unassigned)
|
|
secret VARCHAR(255) NOT NULL, -- per-device HMAC key; rotate via admin CLI, never log
|
|
config_version INT UNSIGNED NOT NULL DEFAULT 1, -- authoritative version served by GET /api/config
|
|
last_fps DOUBLE NULL,
|
|
last_temp DOUBLE NULL,
|
|
last_throttle TINYINT(1) NULL,
|
|
reported_config_version INT UNSIGNED NULL, -- version the device last claimed via heartbeat
|
|
last_heartbeat_at BIGINT UNSIGNED NULL,
|
|
created_at BIGINT UNSIGNED NOT NULL DEFAULT (UNIX_TIMESTAMP()),
|
|
PRIMARY KEY (device_id),
|
|
KEY ix_devices_site (site)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- cameras: per-device camera inventory (design doc §7.4). No dedicated
|
|
-- PdoCameraRepository exists in P1-B -- zones/events reference camera_id
|
|
-- directly as a free-form string. Kept for documentation/future per-camera
|
|
-- config and to avoid a breaking migration when that repository is added.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_cameras (
|
|
camera_id VARCHAR(64) NOT NULL,
|
|
device_id VARCHAR(64) NOT NULL,
|
|
name VARCHAR(190) NULL,
|
|
role VARCHAR(32) NULL, -- app-whitelisted: fall_critical|zone|low
|
|
created_at BIGINT UNSIGNED NOT NULL DEFAULT (UNIX_TIMESTAMP()),
|
|
PRIMARY KEY (camera_id),
|
|
KEY ix_cameras_device (device_id),
|
|
CONSTRAINT chk_cameras_role CHECK (role IS NULL OR role IN ('fall_critical','zone','low'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- zones: polygon regions per camera, edited via the zone-config dashboard.
|
|
-- No FK to cameras/devices: zones may legitimately be created before a
|
|
-- `cameras` row exists for a given camera_id (no camera-registration flow in
|
|
-- P1-B), so a FK here would spuriously reject valid app writes.
|
|
-- ---------------------------------------------------------------------------
|
|
-- ---------------------------------------------------------------------------
|
|
-- Improvement action log (개선 조치): a site fix ("미끄럼 매트 설치") recorded against a
|
|
-- camera; the report compares detection density before/after to prove the fix worked.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_actions (
|
|
action_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
camera_id VARCHAR(64) NOT NULL,
|
|
site VARCHAR(64) NULL, -- access-control scope, resolved from the camera's events at insert
|
|
event_id INT UNSIGNED NULL, -- 이 조치의 계기가 된 사건 (상세 화면에서 바로 기록할 때)
|
|
note VARCHAR(500) NOT NULL,
|
|
taken_at BIGINT UNSIGNED NOT NULL, -- when the fix went in (epoch s)
|
|
created_by VARCHAR(190) NOT NULL,
|
|
created_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (action_id),
|
|
KEY ix_actions_taken (taken_at),
|
|
KEY ix_actions_site_taken (site, taken_at),
|
|
KEY ix_actions_event (event_id)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
CREATE TABLE move_zones (
|
|
zone_id CHAR(32) NOT NULL, -- bin2hex(random_bytes(16)), same convention as snapshot uuids
|
|
device_id VARCHAR(64) NOT NULL,
|
|
camera_id VARCHAR(64) NOT NULL,
|
|
level VARCHAR(16) NOT NULL, -- app-whitelisted: warning|restricted|pinch (ZoneConfigController::LEVELS)
|
|
polygon_json JSON NOT NULL,
|
|
rules_json JSON NULL, -- optional occupancy rules: {"min_workers","max_workers","max_dwell_s"} (ZoneConfigController validates)
|
|
updated_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (zone_id),
|
|
KEY ix_zones_camera (camera_id),
|
|
KEY ix_zones_device (device_id),
|
|
CONSTRAINT chk_zones_level CHECK (level IN ('warning','restricted','pinch'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- events: ingested detections. event_type/severity are already validated in
|
|
-- App\Ingest\EventValidator before insert(); the CHECK constraints below are
|
|
-- defense-in-depth only, not the primary validation layer.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_events (
|
|
event_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
event_uuid CHAR(32) NULL, -- edge-generated idempotency key (dedup a re-sent event); NULL for legacy rows
|
|
device_id VARCHAR(64) NOT NULL,
|
|
site VARCHAR(64) NULL, -- denormalized from move_devices.site at ingest; access-control scope
|
|
camera_id VARCHAR(64) NOT NULL,
|
|
event_type VARCHAR(32) NOT NULL,
|
|
severity VARCHAR(16) NOT NULL,
|
|
track_id INT NULL,
|
|
zone_id CHAR(32) NULL,
|
|
occurred_at DOUBLE NOT NULL, -- device-reported epoch seconds (float)
|
|
duration_s DOUBLE NOT NULL DEFAULT 0,
|
|
metrics_json JSON NULL,
|
|
spatial_x DOUBLE NULL, -- normalized foot/reference point (0..1 of the frame),
|
|
spatial_y DOUBLE NULL, -- heatmap raw material; NULL for legacy/deviceless events
|
|
acknowledged_by VARCHAR(190) NULL,
|
|
acknowledged_at BIGINT UNSIGNED NULL,
|
|
false_positive_by VARCHAR(190) NULL, -- operator's 오탐 tag (per-camera FP-rate stats; no auto-tuning)
|
|
false_positive_at BIGINT UNSIGNED NULL,
|
|
-- 증거 보존 보류. 이 시각까지는 보존기간 정리가 이 사건의 스냅샷을 지우지 않는다.
|
|
-- 중대재해 조사는 몇 주~몇 달인데 스냅샷 기본 보존은 7일이라, 보류가 없으면 "사고 후
|
|
-- 원인 분석"이라는 목적 자체가 8일째에 사라진다. 자동 연장이 아니라 명시적 조작으로만
|
|
-- 걸리고(자동이면 결국 전량 보존이 되어 프라이버시 약속이 무너진다) 감사에 남는다.
|
|
hold_until BIGINT UNSIGNED NULL,
|
|
received_at BIGINT UNSIGNED NOT NULL, -- server receipt time (PHP time()), for clock-skew flagging (design doc §8)
|
|
PRIMARY KEY (event_id),
|
|
-- Idempotency: a UNIQUE key on event_uuid makes a re-sent event (POST succeeded server-side
|
|
-- but the ACK timed out client-side) collide instead of duplicating. MySQL allows multiple
|
|
-- NULLs in a UNIQUE index, so legacy rows without a uuid do not conflict.
|
|
UNIQUE KEY uq_events_uuid (event_uuid),
|
|
KEY ix_events_occurred_at (occurred_at),
|
|
KEY ix_events_severity_occurred_at (severity, occurred_at),
|
|
KEY ix_events_site_occurred_at (site, occurred_at), -- site-scoped dashboard list (P3-2b)
|
|
KEY ix_events_hold (hold_until), -- 보존 정리가 보류된 건만 빠르게 걸러낸다
|
|
KEY ix_events_device (device_id),
|
|
KEY ix_events_camera (camera_id, event_id), -- zone editor: newest snapshot per camera
|
|
CONSTRAINT chk_events_type CHECK (event_type IN (
|
|
'fall','fall_from_height','no_motion','immobility','ghost','zone_warning',
|
|
'zone_restricted','pinch','monitoring_degraded','device_fault',
|
|
'lone_worker','overcapacity','overstay','near_fall','ergo_report',
|
|
-- 감시 자체의 건전성: 카메라가 틀어지거나 가려지면 존 판정이 조용히 무의미해진다.
|
|
'camera_moved','camera_obstructed','low_light',
|
|
-- 아무도 확인하지 않았다는 사실 자체를 사건으로 만든다 (에스컬레이션 디스패처가 발행).
|
|
'unacked_escalation','multi_down',
|
|
-- 모델 확보 후 활성화되는 자리. 엣지 감지기는 있으나 기본 비활성이다.
|
|
'ppe_missing','proximity','fire_smoke'
|
|
)),
|
|
CONSTRAINT chk_events_severity CHECK (severity IN ('critical','high','warning','info'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- event_snapshots: re-encoded JPEG references. Files live under storage/
|
|
-- (webroot-adjacent, served only via the authenticated SnapshotProxy). FK to
|
|
-- events is safe because IngestController always inserts the parent event row
|
|
-- before calling SnapshotRepository::add() for its snapshots.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_event_snapshots (
|
|
snapshot_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
event_id INT UNSIGNED NOT NULL,
|
|
file_uuid CHAR(32) NOT NULL, -- bin2hex(random_bytes(16)); on-disk filename = <uuid>.jpg
|
|
content_hash CHAR(64) NOT NULL, -- sha256 of the re-encoded bytes
|
|
created_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (snapshot_id),
|
|
UNIQUE KEY uq_event_snapshots_uuid (file_uuid),
|
|
-- Idempotent snapshot recovery: a re-sent event carries the SAME re-encoded bytes (same
|
|
-- content_hash), so this makes a duplicate/racing store collide instead of writing a second
|
|
-- copy. Lets the dedup path safely re-store missing evidence without duplication.
|
|
UNIQUE KEY uq_event_snapshots_event_hash (event_id, content_hash),
|
|
KEY ix_event_snapshots_event (event_id),
|
|
KEY ix_event_snapshots_created (created_at), -- retention purge scans WHERE created_at < ?
|
|
CONSTRAINT fk_event_snapshots_event FOREIGN KEY (event_id) REFERENCES move_events (event_id)
|
|
ON DELETE CASCADE
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- thresholds: per-device (optionally per-camera/zone) tunables synced via
|
|
-- GET /api/config. No PdoThresholdRepository write method exists in P1-B --
|
|
-- rows are seeded/edited directly (phpMyAdmin) until a management UI lands.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_thresholds (
|
|
id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
device_id VARCHAR(64) NOT NULL,
|
|
camera_id VARCHAR(64) NULL,
|
|
zone_id CHAR(32) NULL,
|
|
`key` VARCHAR(64) NOT NULL,
|
|
`value` VARCHAR(255) NOT NULL,
|
|
updated_by VARCHAR(190) NULL, -- 편집 UI(관리자 전용)가 남기는 감사 흔적
|
|
updated_at BIGINT UNSIGNED NULL,
|
|
PRIMARY KEY (id),
|
|
KEY ix_thresholds_device (device_id),
|
|
UNIQUE KEY uq_thresholds_scope_key (device_id, camera_id, zone_id, `key`)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- nonces: replay-protection store for ingest/config/heartbeat HMAC requests.
|
|
-- `nonce` is the PRIMARY KEY (unique) so PdoNonceRepository::remember() is
|
|
-- safe against a duplicate-insert race: IngestController already calls
|
|
-- seen() before remember(), and a thrown duplicate-key error on a racing
|
|
-- remember() is an acceptable, expected outcome (closes the seen()/remember()
|
|
-- TOCTOU flagged in the Task 7 review -- the PK uniqueness is the real guard,
|
|
-- not the earlier seen() check).
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_nonces (
|
|
nonce VARCHAR(190) NOT NULL,
|
|
expires_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (nonce),
|
|
KEY ix_nonces_expires_at (expires_at) -- for a periodic `DELETE FROM move_nonces WHERE expires_at < ?` cleanup job
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- rate_limits: per-key sliding-window counters for ingest rate limiting. The
|
|
-- (rate_key, window_start) primary key backs an atomic
|
|
-- `INSERT ... ON DUPLICATE KEY UPDATE count = count + 1` upsert in
|
|
-- PdoRateLimiterRepository::hit() -- no read-modify-write race window.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_rate_limits (
|
|
rate_key VARCHAR(190) NOT NULL,
|
|
window_start BIGINT UNSIGNED NOT NULL, -- floor(now / 60): the minute bucket
|
|
count INT UNSIGNED NOT NULL DEFAULT 0,
|
|
PRIMARY KEY (rate_key, window_start),
|
|
-- The reap `DELETE ... WHERE window_start < ?` cannot use the PK (window_start is its trailing
|
|
-- column), so give it its own index to avoid a locking full-scan on the pre-auth hot path.
|
|
KEY ix_rate_limits_window (window_start)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- audit_log: append-only action log (acknowledge / zone save+delete / login /
|
|
-- threshold edit / suppression / handover).
|
|
--
|
|
-- TAMPER-EVIDENT: each row carries prev_hash (the previous row's entry_hash) and
|
|
-- entry_hash = sha256(id|actor|action|target|before|after|ts|prev_hash). Editing or
|
|
-- deleting a row breaks every hash after it, so a doctored log is DETECTABLE even
|
|
-- though shared hosting cannot grant append-only rights at the DB level. Verified by
|
|
-- `php tools/verify_audit_chain.php`. Rows written before the chain existed keep NULL
|
|
-- hashes and the verifier reports them as the pre-chain segment rather than as damage.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_audit_log (
|
|
id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
actor VARCHAR(190) NOT NULL,
|
|
action VARCHAR(64) NOT NULL,
|
|
target VARCHAR(190) NOT NULL,
|
|
before_json JSON NOT NULL,
|
|
after_json JSON NOT NULL,
|
|
ts BIGINT UNSIGNED NOT NULL,
|
|
prev_hash CHAR(64) NULL,
|
|
entry_hash CHAR(64) NULL,
|
|
PRIMARY KEY (id),
|
|
KEY ix_audit_log_ts (ts),
|
|
KEY ix_audit_log_actor (actor),
|
|
KEY ix_audit_log_entry_hash (entry_hash)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- suppressions: planned-work alarm suppression (정비 모드 / 작업허가).
|
|
-- Detection is NOT paused -- events are still ingested and still appear in the console
|
|
-- and in reports. Only NOTIFICATION is withheld, and every suppression carries a
|
|
-- mandatory reason, a hard expiry (no open-ended suppression) and an audit entry.
|
|
-- Rationale: the single largest source of false alarms is planned maintenance, and a
|
|
-- console that cries wolf during every shutdown is a console the site switches off.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_suppressions (
|
|
suppression_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
scope_type VARCHAR(16) NOT NULL, -- device | camera | zone
|
|
scope_id VARCHAR(64) NOT NULL,
|
|
site VARCHAR(64) NULL,
|
|
reason VARCHAR(500) NOT NULL,
|
|
starts_at BIGINT UNSIGNED NOT NULL,
|
|
ends_at BIGINT UNSIGNED NOT NULL,
|
|
created_by VARCHAR(190) NOT NULL,
|
|
created_at BIGINT UNSIGNED NOT NULL,
|
|
revoked_by VARCHAR(190) NULL,
|
|
revoked_at BIGINT UNSIGNED NULL,
|
|
PRIMARY KEY (suppression_id),
|
|
KEY ix_suppressions_window (ends_at, starts_at),
|
|
KEY ix_suppressions_scope (scope_type, scope_id),
|
|
KEY ix_suppressions_site (site),
|
|
CONSTRAINT chk_suppressions_scope CHECK (scope_type IN ('device','camera','zone'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- alert_channels / alert_log: how a detection reaches a human.
|
|
-- Air-gapped sites cannot use FCM/Slack, so the channels are the ones a closed network
|
|
-- actually has: `local` (the edge node's own beacon/buzzer over GPIO -- the only path
|
|
-- that needs no network at all), `email` (in-house SMTP relay) and `webhook` (in-house
|
|
-- groupware / SMS gateway). step>0 channels fire only after escalate_after_s of the
|
|
-- event still being unacknowledged, which is what turns "we detected it" into "someone
|
|
-- was made to look at it".
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_alert_channels (
|
|
channel_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
site VARCHAR(64) NULL,
|
|
kind VARCHAR(16) NOT NULL,
|
|
name VARCHAR(190) NOT NULL,
|
|
config_json JSON NOT NULL, -- recipients / url / gpio pin. NEVER credentials.
|
|
min_severity VARCHAR(16) NOT NULL DEFAULT 'high',
|
|
enabled TINYINT(1) NOT NULL DEFAULT 1,
|
|
step INT UNSIGNED NOT NULL DEFAULT 0,
|
|
escalate_after_s INT UNSIGNED NOT NULL DEFAULT 0,
|
|
created_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (channel_id),
|
|
KEY ix_alert_channels_site (site, enabled),
|
|
CONSTRAINT chk_alert_channels_kind CHECK (kind IN ('local','email','webhook')),
|
|
CONSTRAINT chk_alert_channels_sev CHECK (min_severity IN ('critical','high','warning','info'))
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
CREATE TABLE move_alert_log (
|
|
alert_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
event_id INT UNSIGNED NOT NULL,
|
|
channel_id INT UNSIGNED NOT NULL,
|
|
status VARCHAR(16) NOT NULL,
|
|
attempts INT UNSIGNED NOT NULL DEFAULT 0,
|
|
last_error VARCHAR(500) NULL,
|
|
queued_at BIGINT UNSIGNED NOT NULL,
|
|
sent_at BIGINT UNSIGNED NULL,
|
|
PRIMARY KEY (alert_id),
|
|
-- The dispatcher runs from cron and overlapping runs are expected; this unique key is
|
|
-- the only thing standing between a site and the same fall being phoned in five times.
|
|
UNIQUE KEY uq_alert_log_event_channel (event_id, channel_id),
|
|
KEY ix_alert_log_status (status, queued_at),
|
|
CONSTRAINT chk_alert_log_status CHECK (status IN ('pending','sent','failed','suppressed')),
|
|
CONSTRAINT fk_alert_log_event FOREIGN KEY (event_id) REFERENCES move_events (event_id) ON DELETE CASCADE
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- shift_handovers: who took responsibility for what, and when.
|
|
-- The 24h brief tells an operator what happened; it does not record that anyone
|
|
-- accepted the backlog. Without that, unacknowledged events evaporate across a shift
|
|
-- change with nobody accountable for them.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_shift_handovers (
|
|
handover_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
site VARCHAR(64) NULL,
|
|
from_user VARCHAR(190) NULL,
|
|
to_user VARCHAR(190) NOT NULL,
|
|
period_start BIGINT UNSIGNED NOT NULL,
|
|
period_end BIGINT UNSIGNED NOT NULL,
|
|
unacked_count INT UNSIGNED NOT NULL,
|
|
severe_count INT UNSIGNED NOT NULL,
|
|
note VARCHAR(1000) NULL,
|
|
created_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (handover_id),
|
|
KEY ix_handovers_site_created (site, created_at)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- risk_controls: 위험성평가 연계. One row per (site, detection type): the hazard it
|
|
-- stands for, the control currently in place, and the 5x5 rating. This is what lets the
|
|
-- system print "what are we controlling and how well", which is the artefact a safety
|
|
-- management system is actually audited on.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_risk_controls (
|
|
control_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
site VARCHAR(64) NULL,
|
|
event_type VARCHAR(32) NOT NULL,
|
|
hazard VARCHAR(300) NOT NULL,
|
|
control VARCHAR(500) NOT NULL,
|
|
severity_rating TINYINT UNSIGNED NOT NULL,
|
|
likelihood_rating TINYINT UNSIGNED NOT NULL,
|
|
reviewed_by VARCHAR(190) NULL,
|
|
reviewed_at BIGINT UNSIGNED NULL,
|
|
updated_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (control_id),
|
|
UNIQUE KEY uq_risk_controls_site_type (site, event_type),
|
|
CONSTRAINT chk_risk_sev CHECK (severity_rating BETWEEN 1 AND 5),
|
|
CONSTRAINT chk_risk_like CHECK (likelihood_rating BETWEEN 1 AND 5)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ---------------------------------------------------------------------------
|
|
-- device_metrics: heartbeat telemetry as a TIME SERIES.
|
|
-- move_devices holds only the latest value, so a node whose inference latency doubled
|
|
-- overnight looks identical to a healthy one. Appending each heartbeat is what makes
|
|
-- slow degradation visible before it becomes a monitoring outage.
|
|
-- ---------------------------------------------------------------------------
|
|
CREATE TABLE move_device_metrics (
|
|
metric_id INT UNSIGNED NOT NULL AUTO_INCREMENT,
|
|
device_id VARCHAR(64) NOT NULL,
|
|
ts BIGINT UNSIGNED NOT NULL,
|
|
fps DOUBLE NULL,
|
|
temp DOUBLE NULL,
|
|
throttle TINYINT(1) NULL,
|
|
queue_pending INT UNSIGNED NULL,
|
|
cameras_total INT UNSIGNED NULL,
|
|
cameras_erroring INT UNSIGNED NULL,
|
|
errors INT UNSIGNED NULL,
|
|
PRIMARY KEY (metric_id),
|
|
KEY ix_device_metrics_device_ts (device_id, ts)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- 감시자의 생존 기록.
|
|
--
|
|
-- 죽은 노드는 스스로 신고할 수 없어서 크론(tools/dispatch_alerts.php)이 하트비트의 부재로
|
|
-- 판정한다. 그런데 그 크론이 멈추면 아무도 판정하지 않는다 — 통보·5분 확대·장치 이상 발행이
|
|
-- 전부 그 한 프로세스에 걸려 있어서, 멈춘 순간부터 화면은 조용해지고 그 침묵은 "사고 없음"과
|
|
-- 구별되지 않는다. 감시자의 생존은 감시자가 아닌 쪽이 읽어야 하므로, 크론이 매 실행마다 자기
|
|
-- 시각을 여기 남기고 대시보드(/api/devices)가 그 나이를 본다.
|
|
--
|
|
-- 범용 키-값이지만 용도를 넓히지 말 것: 설정은 config.php, 임계값은 move_thresholds 가 있다.
|
|
-- 여기 들어갈 것은 "프로세스가 언제 살아 있었는가" 뿐이다.
|
|
CREATE TABLE move_app_state (
|
|
k VARCHAR(64) NOT NULL,
|
|
v VARCHAR(255) NOT NULL,
|
|
updated_at BIGINT UNSIGNED NOT NULL,
|
|
PRIMARY KEY (k)
|
|
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4;
|
|
|
|
-- ===========================================================================
|
|
-- UPGRADING AN EXISTING DATABASE (fresh installs already have everything above):
|
|
-- do NOT copy the statements below by hand — they are SQL comments and will not
|
|
-- run. Apply the RUNNABLE migration instead:
|
|
--
|
|
-- cloud/migrations/2026-07-17-p1-9-p3-2b.sql
|
|
-- cloud/migrations/2026-07-20-move-table-prefix.sql (run this one LAST)
|
|
--
|
|
-- The second migration renames every table to the `move_` prefix used above and
|
|
-- by App\Repository\Pdo\* -- an existing DB must be renamed before deploy or all
|
|
-- queries fail with "table doesn't exist".
|
|
--
|
|
-- It adds devices.site / events.site / events.event_uuid, the retention and
|
|
-- rate-limit reap indexes, AND backfills events.site from move_devices.site (without
|
|
-- that backfill, historical events vanish for a scoped operator). Run it once in
|
|
-- phpMyAdmin BEFORE deploying code that reads these columns, then provision:
|
|
-- UPDATE move_devices SET site = 'factory-a' WHERE device_id = 'pi-1';
|
|
-- UPDATE move_users SET site_scope = 'factory-a' WHERE username = 'operator-a';
|
|
-- A user with site_scope = NULL is a global operator; '' is NOT global (matches
|
|
-- nothing — a blank scope fails closed).
|
|
-- ===========================================================================
|