Datasets:

dicemy
/

DataClawEval

	-- mysql_013 database initialization: create tables + load seed data
	-- Executed via: mysql -u root < init_db.sql

	-- Fix: MySQL root default auth_socket blocks pymysql TCP connections,
	-- switch to mysql_native_password so pymysql (Python) can connect.
	ALTER USER 'root'@'localhost' IDENTIFIED WITH mysql_native_password BY 'root123';
	FLUSH PRIVILEGES;

	CREATE DATABASE IF NOT EXISTS internal_platform_db
	DEFAULT CHARACTER SET utf8mb4
	DEFAULT COLLATE utf8mb4_unicode_ci;

	USE internal_platform_db;

	-- Input table 1: notebook_span_info
	CREATE TABLE IF NOT EXISTS notebook_span_info_mysql_013 (
	databus_imp_date VARCHAR(256) NOT NULL COMMENT '数据导入日期',
	trace_id VARCHAR(256) NOT NULL COMMENT '追踪ID',
	span_kind VARCHAR(256) DEFAULT NULL COMMENT 'Span类型',
	error_code VARCHAR(256) DEFAULT NULL COMMENT '错误码',
	service_name VARCHAR(256) DEFAULT NULL COMMENT '服务名称',
	message VARCHAR(256) DEFAULT NULL COMMENT '消息',
	parent_span_id VARCHAR(256) DEFAULT NULL COMMENT '父Span ID',
	span_name VARCHAR(256) DEFAULT NULL COMMENT 'Span名称',
	duration BIGINT DEFAULT NULL COMMENT '持续时间',
	span_id VARCHAR(256) DEFAULT NULL COMMENT 'Span ID',
	space_id VARCHAR(256) DEFAULT NULL COMMENT '空间ID',
	span_type VARCHAR(256) DEFAULT NULL COMMENT 'Span类型',
	start_time VARCHAR(256) DEFAULT NULL COMMENT '开始时间(毫秒时间戳)',
	end_time VARCHAR(256) DEFAULT NULL COMMENT '结束时间(毫秒时间戳)',
	service_instance VARCHAR(256) DEFAULT NULL COMMENT '服务实例',
	status_code INT DEFAULT NULL COMMENT '状态码',
	datawd_project_id VARCHAR(256) DEFAULT NULL COMMENT '项目ID',
	datawd_task_id VARCHAR(256) DEFAULT NULL COMMENT '任务ID',
	datawd_task_instance_id VARCHAR(256) DEFAULT NULL COMMENT '任务实例ID',
	compute_type VARCHAR(256) DEFAULT NULL COMMENT '计算类型'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- Input table 2: dwd_gputj_service_instance_map
	CREATE TABLE IF NOT EXISTS dwd_gputj_service_instance_map_mysql_013 (
	dt VARCHAR(256) NOT NULL COMMENT '日期',
	service_name VARCHAR(256) DEFAULT NULL COMMENT '服务名称',
	instance_uuid VARCHAR(256) NOT NULL COMMENT '实例UUID',
	service_id BIGINT DEFAULT NULL COMMENT '服务ID',
	version_id BIGINT DEFAULT NULL COMMENT '版本ID',
	`desc` VARCHAR(256) DEFAULT NULL COMMENT '描述',
	create_time DOUBLE DEFAULT NULL COMMENT '创建时间',
	update_time DOUBLE DEFAULT NULL COMMENT '更新时间',
	workload_name VARCHAR(256) DEFAULT NULL COMMENT '工作负载名称'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- Input table 3: dwd_ml_platform_instance_podname
	CREATE TABLE IF NOT EXISTS dwd_ml_platform_instance_podname_mysql_013 (
	dt VARCHAR(256) NOT NULL COMMENT '日期',
	instance_uuid VARCHAR(256) NOT NULL COMMENT '实例UUID',
	pod_name VARCHAR(256) DEFAULT NULL COMMENT 'Pod名称',
	pod_phase VARCHAR(256) DEFAULT NULL COMMENT 'Pod阶段',
	pod_create_time BIGINT DEFAULT NULL COMMENT 'Pod创建时间',
	pod_update_time BIGINT DEFAULT NULL COMMENT 'Pod更新时间',
	namespace VARCHAR(256) DEFAULT NULL COMMENT '命名空间',
	trial_job_name VARCHAR(256) DEFAULT NULL COMMENT '试用任务名称',
	create_time BIGINT DEFAULT NULL COMMENT '创建时间',
	modify_time BIGINT DEFAULT NULL COMMENT '修改时间',
	fix_modify_time BIGINT DEFAULT NULL COMMENT '修正修改时间'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- Input table 4: gputj_gpu_info_parsed_agg_1min
	CREATE TABLE IF NOT EXISTS gputj_gpu_info_parsed_agg_1min_mysql_013 (
	dt VARCHAR(256) NOT NULL COMMENT '日期',
	pod_name VARCHAR(256) DEFAULT NULL COMMENT 'Pod名称',
	pkg_agg_time VARCHAR(256) DEFAULT NULL COMMENT '聚合时间',
	k8s_container_vgpu_gpu_util_sum DOUBLE DEFAULT NULL COMMENT 'GPU利用率总和',
	k8s_container_vgpu_gpu_util_count BIGINT DEFAULT NULL COMMENT 'GPU利用率计数',
	k8s_container_resource_request_gpu_sum DOUBLE DEFAULT NULL COMMENT 'GPU申请量总和',
	k8s_container_resource_request_gpu_count BIGINT DEFAULT NULL COMMENT 'GPU申请量计数'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- Input table 5: notebook_engine_info
	CREATE TABLE IF NOT EXISTS notebook_engine_info_mysql_013 (
	databus_imp_date VARCHAR(256) NOT NULL COMMENT '数据导入日期',
	trace_id VARCHAR(256) NOT NULL COMMENT '追踪ID',
	service_name VARCHAR(256) DEFAULT NULL COMMENT '服务名称',
	compute_type VARCHAR(256) DEFAULT NULL COMMENT '计算类型',
	datawd_project_id VARCHAR(256) DEFAULT NULL COMMENT '项目ID',
	datawd_task_id VARCHAR(256) DEFAULT NULL COMMENT '任务ID',
	datawd_task_instance_id VARCHAR(256) DEFAULT NULL COMMENT '任务实例ID',
	is_permanent VARCHAR(256) DEFAULT NULL COMMENT '是否永久',
	serving_id VARCHAR(256) DEFAULT NULL COMMENT '服务ID',
	user_name VARCHAR(256) DEFAULT NULL COMMENT '用户名',
	is_model_service VARCHAR(256) DEFAULT NULL COMMENT '是否模型服务',
	compute_info VARCHAR(256) DEFAULT NULL COMMENT '计算信息',
	num_gpu VARCHAR(256) DEFAULT NULL COMMENT 'GPU数量',
	replicas VARCHAR(256) DEFAULT NULL COMMENT '副本数',
	gpu_type VARCHAR(256) DEFAULT NULL COMMENT 'GPU类型',
	`timestamp` VARCHAR(256) DEFAULT NULL COMMENT '时间戳',
	ray_job_id VARCHAR(256) DEFAULT NULL COMMENT 'Ray任务ID',
	reason VARCHAR(256) DEFAULT NULL COMMENT '原因',
	log_url VARCHAR(256) DEFAULT NULL COMMENT '日志URL',
	dashboard_url VARCHAR(256) DEFAULT NULL COMMENT '仪表盘URL'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- Output table: dwd_notebook_instance_pod_cross_day_detail_d_cand
	CREATE TABLE IF NOT EXISTS dwd_notebook_instance_pod_cross_day_detail_d_cand_mysql_013 (
	trace_id VARCHAR(256) NOT NULL COMMENT '追踪ID',
	datawd_project_id VARCHAR(256) DEFAULT NULL COMMENT '项目ID',
	datawd_task_id VARCHAR(256) DEFAULT NULL COMMENT '任务ID',
	datawd_task_instance_id VARCHAR(256) DEFAULT NULL COMMENT '任务实例ID',
	compute_type VARCHAR(256) DEFAULT NULL COMMENT '计算类型',
	status_code INT DEFAULT NULL COMMENT '状态码',
	instance_run_time INT DEFAULT NULL COMMENT '实例运行时长(秒)',
	code_run_time INT DEFAULT NULL COMMENT '代码运行时长(秒)',
	resource_wait_time INT DEFAULT NULL COMMENT '资源等待时长(秒)',
	code_start_time VARCHAR(256) DEFAULT NULL COMMENT '代码开始时间',
	code_end_time VARCHAR(256) DEFAULT NULL COMMENT '代码结束时间',
	instance_start_time VARCHAR(256) DEFAULT NULL COMMENT '实例开始时间',
	instance_end_time VARCHAR(256) DEFAULT NULL COMMENT '实例结束时间',
	serving_id VARCHAR(256) DEFAULT NULL COMMENT '服务ID',
	is_permanent TINYINT(1) DEFAULT NULL COMMENT '是否永久(0/1)',
	apply_for_gpu_count INT DEFAULT NULL COMMENT '申请GPU数量',
	pod_name VARCHAR(256) DEFAULT NULL COMMENT 'Pod名称',
	pkg_agg_time VARCHAR(256) DEFAULT NULL COMMENT 'GPU聚合时间',
	gpu_util DOUBLE DEFAULT NULL COMMENT 'GPU利用率',
	gpu_count DOUBLE DEFAULT NULL COMMENT 'GPU数量',
	p_date VARCHAR(256) DEFAULT NULL COMMENT '日期',
	dt VARCHAR(256) NOT NULL COMMENT '分区日期'
	) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci;

	-- ============================================================
	-- Seed data
	-- ============================================================

	-- notebook_span_info: 9 rows (trace001 with cross-day span, trace002 single-day)
	-- Timestamps are in UTC+8: 1746460800000(+08)=2026-05-06 00:00:00, etc.
	-- service_instance links to dwd_gputj_service_instance_map.instance_uuid
	INSERT INTO notebook_span_info_mysql_013 VALUES
	('2026050700','trace001','','','notebook-runner','','','runner.execute',0,'','','','1777996800000','1778097600000','uuid001',0,'proj01','task01','inst01','ray'),
	('2026050700','trace001','','','notebook-runner','','','execute.code',0,'','','','1778000400000','1778090400000','uuid001',0,'proj01','task01','inst01','ray'),
	-- NOTE: set.permanent.compute for trace001 lives ONLY at databus_imp_date=2026050400 (an
	-- earlier import date inside the [0504,0507] window). The baseline still picks it up, so
	-- resource_wait_time=600. This is the 诱饵 for the "outer date window lower bound" rule:
	-- tightening the lower bound to 0507 drops this row -> resource_wait_time becomes NULL.
	('2026050400','trace001','','','notebook-runner','','','runner.execute',0,'','','','1777996800000','1778097600000','uuid001',0,'proj01','task01','inst01','ray'),
	('2026050400','trace001','','','notebook-runner','','','execute.code',0,'','','','1778000400000','1778090400000','uuid001',0,'proj01','task01','inst01','ray'),
	('2026050400','trace001','','','notebook-runner','','','set.permanent.compute',0,'','','','1777996800000','1777997400000','uuid001',0,'proj01','task01','inst01','ray'),
	('2026050700','trace002','','','notebook-runner','','','runner.execute',0,'','','','1778083200000','1778086800000','uuid002',1,'proj02','task02','inst02','ray'),
	('2026050700','trace002','','','notebook-runner','','','client.execute.code',0,'','','','1778083800000','1778086200000','uuid002',1,'proj02','task02','inst02','ray'),
	('2026050700','trace002','','','notebook-runner','','','create.non.permanent.compute',0,'','','','1778083200000','1778083500000','uuid002',1,'proj02','task02','inst02','ray'),
	-- ===== 诱饵 traces for the cross-day anchor filters (anchor subquery on dt=2026050700) =====
	-- trace_d1: ray + notebook-runner but NO runner.execute span -> excluded by the
	-- "span_name='runner.execute'" anchor condition. If that condition is dropped, it leaks in.
	('2026050700','trace_d1','','','notebook-runner','','','execute.code',0,'','','','1778083800000','1778086200000','uuidd1',0,'projd1','taskd1','instd1','ray'),
	-- trace_d2: has runner.execute + notebook-runner but compute_type='spark' -> excluded by the
	-- anchor "compute_type='ray'" condition. Dropping it leaks this trace in.
	('2026050700','trace_d2','','','notebook-runner','','','runner.execute',0,'','','','1778083200000','1778086800000','uuidd2',0,'projd2','taskd2','instd2','spark'),
	-- trace_d3: has runner.execute + ray but service_name='other-svc' -> excluded by the anchor
	-- "service_name='notebook-runner'" condition. Dropping it leaks this trace in.
	('2026050700','trace_d3','','','other-svc','','','runner.execute',0,'','','','1778083200000','1778086800000','uuidd3',0,'projd3','taskd3','instd3','ray');

	-- dwd_gputj_service_instance_map: 2 valid rows + 1 stale-dt 诱饵
	INSERT INTO dwd_gputj_service_instance_map_mysql_013 VALUES
	('2026050700','svc1','uuid001',12345,1,'desc1',0.0,0.0,'wl1'),
	('2026050700','svc2','uuid002',67890,1,'desc2',0.0,0.0,'wl2'),
	-- 诱饵 (map dt filter): stale dt maps service_id 12345 to uuid002 (=> pod-xyz-002). The
	-- gputj_dim t1.dt filter keeps only dt='2026050700', so this is excluded in baseline.
	-- If that dt filter is dropped, trace001 (serving 12345) also joins pod-xyz-002 -> extra rows.
	('2025010100','svc1stale','uuid002',12345,1,'descstale',0.0,0.0,'wlstale');

	-- dwd_ml_platform_instance_podname: 2 valid rows + 1 stale-dt 诱饵
	INSERT INTO dwd_ml_platform_instance_podname_mysql_013 VALUES
	('2026050700','uuid001','pod-abc-001','Running',0,0,'ns1','job1',0,0,0),
	('2026050700','uuid002','pod-xyz-002','Running',0,0,'ns2','job2',0,0,0),
	-- 诱饵 (podname dt filter): stale dt maps uuid001 to an extra pod. The gputj_dim t2.dt filter
	-- keeps only dt='2026050700', so excluded in baseline. If dropped, uuid001 -> 2 pods -> extra rows.
	('2025010100','uuid001','pod-stale-001','Running',0,0,'ns1','job1',0,0,0);

	-- gputj_gpu_info_parsed_agg_1min: 2 valid rows + 2 诱饵 (count>0 filter, time-window join)
	INSERT INTO gputj_gpu_info_parsed_agg_1min_mysql_013 VALUES
	('2026050600','pod-abc-001','2026-05-06 10:30:00',150.0,2,3.0,2),
	('2026050700','pod-xyz-002','2026-05-07 00:30:00',80.0,1,1.0,1),
	-- 诱饵 (count>0 filter): pod-abc-001 inside trace001's instance window but both counts=0 ->
	-- excluded by "(util_count>0 OR request_count>0)". If dropped it leaks (gpu_util=NULL) -> extra row.
	('2026050600','pod-abc-001','2026-05-06 11:00:00',0.0,0,0.0,0),
	-- 诱饵 (gpu time-window join): pod-abc-001 with valid counts but pkg_agg_time on 2026-05-04,
	-- BEFORE trace001's instance_start_time (2026-05-06) -> excluded by the time-window join.
	-- If the window predicate is replaced with TRUE, it leaks -> extra row for trace001.
	('2026050400','pod-abc-001','2026-05-04 10:00:00',60.0,2,2.0,2);

	-- notebook_engine_info
	-- trace001: num_gpu=1 * replicas=2 = apply_for_gpu_count=2 (replicas != 1 so the
	-- "apply = replicas*num_gpu" rule is observable: dropping the replicas factor -> 1).
	-- trace002: num_gpu=1 * replicas=1 = 1.
	INSERT INTO notebook_engine_info_mysql_013 VALUES
	('2026050700','trace001','notebook-runner','ray','proj01','task01','inst01','true','12345','user1','false','info1','1','2','V100','ts1','rj1','','',''),
	('2026050700','trace002','notebook-runner','ray','proj02','task02','inst02','false','67890','user2','false','info2','1','1','A100','ts2','rj2','','',''),
	-- 诱饵 (engine compute_type filter): same trace002 but compute_type='spark' -> excluded by
	-- "compute_type='ray'". If that filter is dropped it leaks, changing serving_id/is_permanent/apply.
	('2026050700','trace002','notebook-runner','spark','proj02','task02','inst02','true','99999','userX','false','infoX','3','2','A100','tsX','rjX','','',''),
	-- 诱饵 (engine service_name filter): same trace002 but service_name='other-svc' compute_type='ray'
	-- -> excluded by "service_name='notebook-runner'". Dropping that filter leaks it.
	('2026050700','trace002','other-svc','ray','proj02','task02','inst02','true','99999','userY','false','infoY','3','2','A100','tsY','rjY','','','');