ClickHouse常用DDL

1 创建数据表与Kafka消费

sql 复制代码

-- 数据表
CREATE TABLE IF NOT EXISTS check_in_log ON CLUSTER ${cluster}
(
    id String DEFAULT generateUUIDv4() COMMENT 'ID',
    check_in_time        DateTime DEFAULT now() COMMENT '签到时间',
    user_id              Int64    DEFAULT -1    COMMENT '用户ID',
    user_ip_addr         String   DEFAULT ''    COMMENT '用户IP地址'
) ENGINE = ReplicatedReplacingMergeTree('/clickhouse/tables/{shard}/{database}/{table}', '{replica}')
PARTITION BY toYYYYMM(check_in_time)  -- 按月分区
ORDER BY (check_in_time, id)          -- 唯一键
TTL check_in_time + INTERVAL 1 YEAR   -- 保留1年
COMMENT '签到记录';

-- Kafka引擎表
CREATE TABLE IF NOT EXISTS check_in_log_stream ON CLUSTER ${cluster}
(
    check_in_time        DateTime DEFAULT now() COMMENT '签到时间',
    user_id              Int64    DEFAULT -1    COMMENT '用户ID',
    user_ip_addr         String   DEFAULT ''    COMMENT '用户IP地址'
) ENGINE = Kafka
SETTINGS
    kafka_broker_list = '${kafkaServers}',
    kafka_topic_list = 'check_in_log',
    kafka_group_name = 'check_in_log_ck_consumer',
    kafka_format = 'JSONEachRow',
    kafka_skip_broken_messages = 1
COMMENT '签到记录Kafka消费';
 
-- Kafka消费表
CREATE MATERIALIZED VIEW IF NOT EXISTS check_in_log_view ON CLUSTER ${cluster}
TO check_in_log AS
SELECT generateUUIDv4() as id, * FROM check_in_log_stream;
 
-- 聚合表
CREATE TABLE IF NOT EXISTS check_in_log_all ON CLUSTER ${cluster}
ENGINE = Distributed(${cluster}, currentDatabase(), check_in_log, rand())
COMMENT '通用审计日志聚合表';

2 修改表

sql 复制代码

ALTER TABLE check_in_log ADD COLUMN IF NOT EXISTS user_name String DEFAULT '' COMMENT '用户名';

3 表迁移

步骤1：创建新表

sql 复制代码

CREATE TABLE IF NOT EXISTS check_in_log_new ON CLUSTER ${cluster}
(
    id String DEFAULT generateUUIDv4() COMMENT 'ID',
    check_in_time        DateTime DEFAULT now() COMMENT '签到时间',
    user_id              Int64    DEFAULT -1    COMMENT '用户ID',
    user_ip_addr         String   DEFAULT ''    COMMENT '用户IP地址'
) ENGINE = ReplicatedReplacingMergeTree('/clickhouse/tables/{shard}/{database}/{table}', '{replica}')
ORDER BY (check_in_time, id)          -- 唯一键
TTL check_in_time + INTERVAL 1 YEAR   -- 保留1年
COMMENT '签到记录';

步骤2：迁移历史数据（全量插入）

sql 复制代码

INSERT INTO check_in_log_new 
SELECT * 
FROM check_in_log;
WHERE check_in_time >= '2023-01-01' AND check_in_time < '2024-01-01';  -- 按年分批

步骤3：检查数据完整性

sql 复制代码

SELECT count() FROM check_in_log;            -- 旧表行数
SELECT count() FROM check_in_log_new;        -- 新表行数（应相同）

步骤4：切换新旧表（原子操作）

sql 复制代码

-- 1. 重命名旧表（备份）
RENAME TABLE check_in_log TO check_in_log_old ON CLUSTER ${cluster};

-- 2. 重命名新表为正式表名
RENAME TABLE check_in_log_new TO check_in_log ON CLUSTER ${cluster};

-- 3. 删除旧表
DROP TABLE check_in_log_old ON CLUSTER ${cluster};