聊聊redis的事务操作

序

本文主要研究一下redis的事务操作

命令

multi与exec

• 命令行

127.0.0.1:6379> multi
OK
127.0.0.1:6379> incr total
QUEUED
127.0.0.1:6379> incr len
QUEUED
127.0.0.1:6379> exec
1) (integer) 2
2) (integer) 2
127.0.0.1:6379> get total
"2"
127.0.0.1:6379> get len
"2"

• lettuce实例

    @Test
    public void testMultiExec(){
        RedisClient client = RedisClient.create("redis://192.168.99.100:6379/0");
        StatefulRedisConnection<String, String> connection = client.connect();
        RedisCommands<String, String> syncCommands = connection.sync();

        syncCommands.set("hello","1");
        syncCommands.set("world","2");

        syncCommands.multi();
        syncCommands.incr("hello");
        syncCommands.incr("world");

        //DefaultTransactionResult[wasRolledBack=false,result=[1, 2, 1, 3, 1]]
        TransactionResult transactionResult = syncCommands.exec();
        System.out.println(transactionResult);
        System.out.println(syncCommands.get("hello"));
        System.out.println(syncCommands.get("world"));
    }

部分执行

• 命令行

127.0.0.1:6379> multi
OK
127.0.0.1:6379> set a hello
QUEUED
127.0.0.1:6379> set b world
QUEUED
127.0.0.1:6379> incr a
QUEUED
127.0.0.1:6379> set c part
QUEUED
127.0.0.1:6379> exec
1) OK
2) OK
3) (error) ERR value is not an integer or out of range
4) OK
127.0.0.1:6379> get a
"hello"
127.0.0.1:6379> get b
"world"
127.0.0.1:6379> get c
"part"

• lettuce实例

    @Test
    public void testMultiExecError(){
        RedisClient client = RedisClient.create("redis://192.168.99.100:6379/0");
        StatefulRedisConnection<String, String> connection = client.connect();
        RedisCommands<String, String> syncCommands = connection.sync();

        syncCommands.multi();
        syncCommands.set("a","hello");
        syncCommands.set("b","world");
        syncCommands.incr("a");
        syncCommands.set("c","part");
        //DefaultTransactionResult[wasRolledBack=false,result=[OK, OK, io.lettuce.core.RedisCommandExecutionException: ERR value is not an integer or out of range, OK, 1]]
        TransactionResult transactionResult = syncCommands.exec();
        System.out.println(transactionResult);
        System.out.println(syncCommands.get("a"));
        System.out.println(syncCommands.get("b"));
        System.out.println(syncCommands.get("c"));
    }

multi与discard

• 命令行

127.0.0.1:6379> set sum 1
OK
127.0.0.1:6379> multi
OK
127.0.0.1:6379> incr sum
QUEUED
127.0.0.1:6379> discard
OK
127.0.0.1:6379> get sum
"1"

• lettuce实例

    @Test
    public void testMultiDiscard(){
        RedisClient client = RedisClient.create("redis://192.168.99.100:6379/0");
        StatefulRedisConnection<String, String> connection = client.connect();
        RedisCommands<String, String> syncCommands = connection.sync();
        syncCommands.incr("key1");
        syncCommands.multi();
        syncCommands.incr("key1");
        //需要有multi才可以执行discard，成功返回OK
        String result = syncCommands.discard();
        System.out.println(result);
        System.out.println(syncCommands.get("key1"));
    }

check and set

    @Test
    public void testWatch(){
        RedisClient client = RedisClient.create("redis://192.168.99.100:6379/0");
        StatefulRedisConnection<String, String> connection = client.connect();
        RedisCommands<String, String> syncCommands = connection.sync();

        String key = "key";
        syncCommands.watch(key);

        //another connection
        StatefulRedisConnection<String, String> conn2 = client.connect();
        RedisCommands<String, String> syncCommands2 = conn2.sync();
        syncCommands2.set(key, "a");

        syncCommands.multi();
        syncCommands.append(key, "b");
        //DefaultTransactionResult [wasRolledBack=true, responses=0]
        TransactionResult transactionResult = syncCommands.exec();
        System.out.println(transactionResult);

        System.out.println(syncCommands.get(key));
    }

小结

• reids提供multi exec/discard指令，类似open commit/rollback transaction，不过exec遇到类型操作等错误时不会滚，该成功执行的命令还是成功执行，该失败的还是失败
• multi exec保证的是，只要exec命令有执行成功，则事务中一系列的命令都能执行，如果exec因为网络等问题，server端没有接收到，则事务中的一系列命令都不会被执行
• discard需要在有调用multi的前提下才能使用，该命令会清空事务队列等待执行的命令
• redis提供watch指令，可以配合multi exec来使用，可以实现类似数据库的乐观锁的机制，一旦watch的key被其他client有更新，则整个exec操作失败

doc

• redis事务