PostgreSQL锁机制与ALTER TABLE阻塞问题解决方案

1. PostgreSQL锁机制基础解析

PostgreSQL作为一款企业级开源关系数据库，其锁机制设计直接影响着系统的并发性能。与MySQL等数据库不同，PostgreSQL采用了多版本并发控制(MVCC)机制，实现了"读不阻塞写，写不阻塞读"的特性。但在表结构变更等场景下，锁冲突问题仍然会显著影响业务连续性。

PostgreSQL的表级锁分为八个级别，按冲突程度从低到高依次为：

• AccessShareLock（SELECT查询）
• RowShareLock（SELECT FOR UPDATE/SHARE）
• RowExclusiveLock（INSERT/UPDATE/DELETE）
• ShareUpdateExclusiveLock（VACUUM/ANALYZE等维护操作）
• ShareLock（CREATE INDEX非并发模式）
• ShareRowExclusiveLock（某些ALTER TABLE操作）
• ExclusiveLock（REFRESH MATERIALIZED VIEW CONCURRENTLY）
• AccessExclusiveLock（DROP/TRUNCATE等DDL操作）

关键提示：ALTER TABLE操作根据具体子命令不同，可能获取ShareRowExclusiveLock或AccessExclusiveLock。例如添加列操作就需要最高级别的AccessExclusiveLock。

2. ALTER TABLE阻塞场景深度分析

2.1 典型阻塞场景重现

让我们通过实验复现一个典型阻塞场景：

会话1（长事务）：

sql复制BEGIN;
SELECT * FROM users WHERE id=1;  -- 获取AccessShareLock
-- 保持事务不结束

会话2（结构变更）：

sql复制ALTER TABLE users ADD COLUMN phone VARCHAR(20);  -- 需要AccessExclusiveLock
-- 此时会话2会挂起

通过pg_locks视图可以观察到锁等待情况：

sql复制SELECT locktype, relation::regclass, pid, mode, granted 
FROM pg_locks 
WHERE relation = 'users'::regclass;

输出结果将显示：

• 会话1持有AccessShareLock（granted=true）
• 会话2等待AccessExclusiveLock（granted=false）

2.2 锁冲突矩阵详解

下表展示了主要锁类型之间的冲突关系：

3. 生产环境解决方案

3.1 预防性措施

1. 事务时间控制：

sql复制SET statement_timeout = '30s';  -- 设置单条SQL超时
SET lock_timeout = '5s';       -- 设置锁等待超时

2. 维护窗口期：

bash复制# 使用pg_blocking_pids()函数识别阻塞源
psql -c "SELECT pg_terminate_backend(pid) 
         FROM pg_stat_activity 
         WHERE pid IN (SELECT unnest(pg_blocking_pids(pg_backend_pid())));"

3. 锁升级策略：

sql复制BEGIN;
LOCK TABLE users IN SHARE MODE;  -- 先获取低级锁
ALTER TABLE users ADD COLUMN phone VARCHAR(20);  -- 锁升级
COMMIT;

3.2 在线DDL最佳实践

1. 并发创建索引：

sql复制CREATE INDEX CONCURRENTLY users_phone_idx ON users(phone);
-- 比普通CREATE INDEX锁级别更低(ShareUpdateExclusiveLock)

2. 避免长时间事务：

python复制# Django示例：使用atomic装饰器控制事务范围
@transaction.atomic
def update_user(request):
    user = User.objects.select_for_update().get(id=1)
    # 业务处理
    return response

3. 使用事件触发器监控：

sql复制CREATE EVENT TRIGGER abort_long_running_ddl
ON ddl_command_start
WHEN TAG IN ('ALTER TABLE')
EXECUTE FUNCTION abort_long_running_ddl();

CREATE OR REPLACE FUNCTION abort_long_running_ddl()
RETURNS event_trigger AS $$
BEGIN
    IF EXISTS (
        SELECT 1 FROM pg_stat_activity 
        WHERE pid <> pg_backend_pid()
        AND query_start < NOW() - INTERVAL '1 minute'
        AND state = 'active'
    ) THEN
        RAISE EXCEPTION '存在长时间运行的事务，DDL操作已中止';
    END IF;
END;
$$ LANGUAGE plpgsql;

4. 高级排查技巧

4.1 锁监控视图增强

创建自定义锁监控视图：

sql复制CREATE VIEW lock_monitor AS
SELECT blocked_locks.pid AS blocked_pid,
       blocking_locks.pid AS blocking_pid,
       blocked_activity.usename AS blocked_user,
       blocking_activity.usename AS blocking_user,
       blocked_activity.query AS blocked_statement,
       blocking_activity.query AS blocking_statement,
       blocked_activity.application_name AS blocked_application,
       blocking_activity.application_name AS blocking_application,
       now() - blocked_activity.query_start AS blocked_duration,
       now() - blocking_activity.query_start AS blocking_duration
FROM pg_catalog.pg_locks blocked_locks
JOIN pg_catalog.pg_stat_activity blocked_activity ON blocked_activity.pid = blocked_locks.pid
JOIN pg_catalog.pg_locks blocking_locks 
    ON blocking_locks.locktype = blocked_locks.locktype
    AND blocking_locks.DATABASE IS NOT DISTINCT FROM blocked_locks.DATABASE
    AND blocking_locks.relation IS NOT DISTINCT FROM blocked_locks.relation
    AND blocking_locks.page IS NOT DISTINCT FROM blocked_locks.page
    AND blocking_locks.tuple IS NOT DISTINCT FROM blocked_locks.tuple
    AND blocking_locks.virtualxid IS NOT DISTINCT FROM blocked_locks.virtualxid
    AND blocking_locks.transactionid IS NOT DISTINCT FROM blocked_locks.transactionid
    AND blocking_locks.classid IS NOT DISTINCT FROM blocked_locks.classid
    AND blocking_locks.objid IS NOT DISTINCT FROM blocked_locks.objid
    AND blocking_locks.objsubid IS NOT DISTINCT FROM blocked_locks.objsubid
    AND blocking_locks.pid != blocked_locks.pid
JOIN pg_catalog.pg_stat_activity blocking_activity ON blocking_activity.pid = blocking_locks.pid
WHERE NOT blocked_locks.GRANTED;

4.2 自动化处理脚本

bash复制#!/bin/bash
# 自动终止阻塞DDL的事务
BLOCKING_PIDS=$(psql -t -A -c "
    SELECT DISTINCT blocking_locks.pid
    FROM pg_locks blocked_locks
    JOIN pg_stat_activity blocked_activity ON blocked_activity.pid = blocked_locks.pid
    JOIN pg_locks blocking_locks ON blocking_locks.locktype = blocked_locks.locktype
        AND blocking_locks.DATABASE IS NOT DISTINCT FROM blocked_locks.DATABASE
        AND blocking_locks.relation IS NOT DISTINCT FROM blocked_locks.relation
        AND blocking_locks.pid != blocked_locks.pid
    JOIN pg_stat_activity blocking_activity ON blocking_activity.pid = blocking_locks.pid
    WHERE blocked_activity.query LIKE 'ALTER TABLE%'
      AND NOT blocked_locks.GRANTED
      AND blocking_activity.query_start < NOW() - INTERVAL '5 minutes'
")

for PID in $BLOCKING_PIDS; do
    psql -c "SELECT pg_terminate_backend($PID)"
done

5. 替代方案与新技术

5.1 逻辑复制方案

sql复制-- 主库
CREATE PUBLICATION users_pub FOR TABLE users;

-- 从库
CREATE TABLE users_new (LIKE users INCLUDING ALL);
ALTER TABLE users_new ADD COLUMN phone VARCHAR(20);
CREATE SUBSCRIPTION users_sub 
CONNECTION 'host=primary dbname=test' 
PUBLICATION users_pub;

-- 数据同步完成后
BEGIN;
ALTER TABLE users RENAME TO users_old;
ALTER TABLE users_new RENAME TO users;
DROP TABLE users_old;
COMMIT;

5.2 PostgreSQL 12+的改进

1. 快速添加列（非空有默认值）：

sql复制ALTER TABLE users ADD COLUMN status VARCHAR(10) NOT NULL DEFAULT 'active';
-- PG11+会立即填充默认值而不重写表

2. 索引创建进度监控：

sql复制SELECT pid, query, phase, tuples_processed, tuples_total 
FROM pg_stat_progress_create_index;

在实际生产环境中，我们曾遇到一个典型案例：某金融系统在业务高峰期执行ALTER TABLE添加字段，导致核心交易表锁等待超过30分钟。通过分析发现是报表系统长查询持有AccessShareLock所致。最终我们采用以下方案解决：

1. 为报表系统建立专用备库
2. 在维护窗口期执行DDL
3. 设置lock_timeout参数
4. 使用pg_terminate_backend终止非关键阻塞会话