chore: initial import

2026-05-12 17:47:32 +08:00
commit fc54ba84b2
104 changed files with 11575 additions and 0 deletions
--- a/tech/migrations/000001_init_schema.up.sql
+++ b/tech/migrations/000001_init_schema.up.sql
@@ -0,0 +1,180 @@
+-- AI-Ops 初始化 schema
+-- 表前缀 ai_ops_，避免与桥项目表名冲突
+
+-- 1. 告警规则
+CREATE TABLE IF NOT EXISTS ai_ops_rules (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    name VARCHAR(128) NOT NULL,
+    metric_source VARCHAR(64) NOT NULL,
+    metric_name VARCHAR(128) NOT NULL,
+    threshold_type VARCHAR(16) NOT NULL CHECK (threshold_type IN ('>', '<', '=', 'regex')),
+    threshold_value TEXT NOT NULL,
+    duration_min INT NOT NULL DEFAULT 1 CHECK (duration_min >= 1),
+    level VARCHAR(8) NOT NULL CHECK (level IN ('P0', 'P1', 'P2', 'P3')),
+    channel_ids UUID[] NOT NULL DEFAULT '{}',
+    healing_action VARCHAR(32) DEFAULT NULL,
+    healing_config JSONB DEFAULT NULL,
+    is_sandboxed BOOLEAN NOT NULL DEFAULT FALSE,
+    enabled BOOLEAN NOT NULL DEFAULT TRUE,
+    created_by VARCHAR(64) NOT NULL,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    version INT NOT NULL DEFAULT 1,
+    CONSTRAINT uq_rules_name UNIQUE (name)
+);
+
+CREATE INDEX IF NOT EXISTS idx_rules_enabled ON ai_ops_rules(enabled);
+
+-- 2. 告警事件
+CREATE TABLE IF NOT EXISTS ai_ops_alerts (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    rule_id UUID NOT NULL REFERENCES ai_ops_rules(id) ON DELETE CASCADE,
+    level VARCHAR(8) NOT NULL,
+    resource_type VARCHAR(64) NOT NULL,
+    resource_id VARCHAR(128) NOT NULL,
+    current_value TEXT NOT NULL,
+    threshold_value TEXT NOT NULL,
+    status VARCHAR(16) NOT NULL DEFAULT 'triggered'
+        CHECK (status IN ('triggered', 'notified', 'healing', 'resolved', 'escalated', 'acknowledged')),
+    is_aggregated BOOLEAN NOT NULL DEFAULT FALSE,
+    aggregated_count INT DEFAULT 0,
+    parent_alert_id UUID NULL REFERENCES ai_ops_alerts(id) ON DELETE SET NULL,
+    started_at TIMESTAMPTZ NOT NULL,
+    resolved_at TIMESTAMPTZ NULL,
+    acknowledged_by VARCHAR(64) NULL,
+    acknowledged_at TIMESTAMPTZ NULL
+);
+
+CREATE INDEX IF NOT EXISTS idx_alerts_status ON ai_ops_alerts(status);
+CREATE INDEX IF NOT EXISTS idx_alerts_started_at ON ai_ops_alerts(started_at DESC);
+CREATE INDEX IF NOT EXISTS idx_alerts_resource ON ai_ops_alerts(resource_type, resource_id);
+
+-- 3. 自愈执行记录
+CREATE TABLE IF NOT EXISTS ai_ops_healings (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    alert_id UUID NOT NULL REFERENCES ai_ops_alerts(id) ON DELETE CASCADE,
+    action_type VARCHAR(32) NOT NULL
+        CHECK (action_type IN ('switch_route', 'throttle', 'restart_instance', 'invoke_script', 'isolate_node')),
+    config JSONB NOT NULL,
+    status VARCHAR(16) NOT NULL DEFAULT 'pending'
+        CHECK (status IN ('pending', 'succeeded', 'failed', 'rolled_back')),
+    dry_run BOOLEAN NOT NULL DEFAULT FALSE,
+    result_detail JSONB NULL,
+    error_code VARCHAR(16) NULL,
+    started_at TIMESTAMPTZ NOT NULL,
+    completed_at TIMESTAMPTZ NULL
+);
+
+CREATE INDEX IF NOT EXISTS idx_healings_alert ON ai_ops_healings(alert_id);
+
+-- 4. 通知渠道
+CREATE TABLE IF NOT EXISTS ai_ops_channels (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    name VARCHAR(128) NOT NULL,
+    channel_type VARCHAR(32) NOT NULL
+        CHECK (channel_type IN ('webhook', 'email', 'feishu', 'wechat', 'sms')),
+    config JSONB NOT NULL,
+    priority INT NOT NULL DEFAULT 1,
+    enabled BOOLEAN NOT NULL DEFAULT TRUE,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+-- 5. 审计日志（append-only，禁止更新和删除）
+CREATE TABLE IF NOT EXISTS ai_ops_audits (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    tenant_id VARCHAR(64) NOT NULL,
+    object_type VARCHAR(64) NOT NULL,
+    object_id VARCHAR(128) NOT NULL,
+    action VARCHAR(32) NOT NULL
+        CHECK (action IN ('create', 'update', 'delete', 'rollback')),
+    before_state JSONB NULL,
+    after_state JSONB NULL,
+    request_id VARCHAR(64) NOT NULL,
+    result_code VARCHAR(16) NOT NULL,
+    source_ip VARCHAR(45) NOT NULL,
+    actor_id VARCHAR(64) NOT NULL,
+    risk_level VARCHAR(8) NOT NULL DEFAULT 'normal'
+        CHECK (risk_level IN ('normal', 'high', 'critical')),
+    parent_audit_id UUID NULL REFERENCES ai_ops_audits(id) ON DELETE SET NULL,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+CREATE INDEX IF NOT EXISTS idx_audits_tenant_created ON ai_ops_audits(tenant_id, created_at DESC);
+CREATE INDEX IF NOT EXISTS idx_audits_object ON ai_ops_audits(object_type, object_id);
+CREATE INDEX IF NOT EXISTS idx_audits_actor ON ai_ops_audits(actor_id, created_at DESC);
+CREATE INDEX IF NOT EXISTS idx_audits_request ON ai_ops_audits(request_id);
+
+-- 审计日志防篡改触发器（仅允许插入，禁止更新和删除）
+CREATE OR REPLACE FUNCTION ai_ops_audit_readonly()
+RETURNS TRIGGER AS $$
+BEGIN
+    RAISE EXCEPTION 'ai_ops_audits is append-only. Updates and deletes are not allowed.';
+END;
+$$ LANGUAGE plpgsql;
+
+DROP TRIGGER IF EXISTS trg_ai_ops_audits_no_update ON ai_ops_audits;
+CREATE TRIGGER trg_ai_ops_audits_no_update
+    BEFORE UPDATE ON ai_ops_audits
+    FOR EACH ROW
+    EXECUTE FUNCTION ai_ops_audit_readonly();
+
+DROP TRIGGER IF EXISTS trg_ai_ops_audits_no_delete ON ai_ops_audits;
+CREATE TRIGGER trg_ai_ops_audits_no_delete
+    BEFORE DELETE ON ai_ops_audits
+    FOR EACH ROW
+    EXECUTE FUNCTION ai_ops_audit_readonly();
+
+-- 6. 时序指标缓存（降级方案，主存储推荐 Prometheus / VictoriaMetrics）
+CREATE TABLE IF NOT EXISTS ai_ops_metrics (
+    id BIGSERIAL PRIMARY KEY,
+    metric_name VARCHAR(128) NOT NULL,
+    labels JSONB NOT NULL DEFAULT '{}',
+    value DOUBLE PRECISION NOT NULL,
+    recorded_at TIMESTAMPTZ NOT NULL
+);
+
+CREATE INDEX IF NOT EXISTS idx_metrics_name_time ON ai_ops_metrics(metric_name, recorded_at DESC);
+
+-- 分区表（按天分区，自动清理 > 7 天的分区）
+CREATE TABLE IF NOT EXISTS ai_ops_metrics_p (
+    id BIGSERIAL NOT NULL,
+    metric_name VARCHAR(128) NOT NULL,
+    labels JSONB NOT NULL DEFAULT '{}',
+    value DOUBLE PRECISION NOT NULL,
+    recorded_at TIMESTAMPTZ NOT NULL,
+    PRIMARY KEY (id, recorded_at)
+) PARTITION BY RANGE (recorded_at);
+
+-- 创建当前日分区
+CREATE TABLE IF NOT EXISTS ai_ops_metrics_p_default PARTITION OF ai_ops_metrics_p
+    DEFAULT;
+
+-- 启用 pg_partman 扩展后可自动管理分区（建议生产环境使用）
+-- SELECT partman.create_parent('public.ai_ops_metrics_p', 'recorded_at', 'native', 'daily');
+
+-- 7. 配置版本管理（支持审计回滚）
+CREATE TABLE IF NOT EXISTS ai_ops_configs (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    object_type VARCHAR(64) NOT NULL,
+    object_id VARCHAR(128) NOT NULL,
+    config_data JSONB NOT NULL,
+    version INT NOT NULL DEFAULT 1,
+    created_by VARCHAR(64) NOT NULL,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
+    CONSTRAINT uq_configs_object_version UNIQUE (object_type, object_id, version)
+);
+
+CREATE INDEX IF NOT EXISTS idx_configs_object ON ai_ops_configs(object_type, object_id, created_at DESC);
+
+-- 8. 状态快照（自愈回滚用）
+CREATE TABLE IF NOT EXISTS ai_ops_snapshots (
+    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
+    healing_id UUID NOT NULL REFERENCES ai_ops_healings(id) ON DELETE CASCADE,
+    snapshot_type VARCHAR(32) NOT NULL
+        CHECK (snapshot_type IN ('route', 'rate_limit', 'instance', 'script', 'node')),
+    before_state JSONB NOT NULL,
+    after_state JSONB NULL,
+    created_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
+);
+
+CREATE INDEX IF NOT EXISTS idx_snapshots_healing ON ai_ops_snapshots(healing_id);