快捷導(dǎo)航

利用redis lua腳本實現(xiàn)時間窗分布式限流

更新時間：2024年03月26日 09:42:14 作者：DimonHo

Lua是一種輕量小巧的腳本語言,Redis是高性能的key-value內(nèi)存數(shù)據(jù)庫,在部分場景下,是對關(guān)系數(shù)據(jù)庫的良好補充,本文給大家介紹了如何利用redis lua腳本實現(xiàn)時間窗分布式限流,需要的朋友可以參考下

需求背景：

限制某sql在30秒內(nèi)最多只能執(zhí)行3次

需求分析

微服務(wù)分布式部署，既然是分布式限流，首先自然就想到了結(jié)合redis的zset數(shù)據(jù)結(jié)構(gòu)來實現(xiàn)。
分析對zset的操作，有幾個步驟，首先，判斷zset中符合rangeScore的元素個數(shù)是否已經(jīng)達(dá)到閾值，如果未達(dá)到閾值，則add元素，并返回true。如果已達(dá)到閾值，則直接返回false。

代碼實現(xiàn)

首先，我們需要根據(jù)需求編寫一個lua腳本

redis.call('ZREMRANGEBYSCORE', KEYS[1], 0, tonumber(ARGV[3]))
local res = 0
if(redis.call('ZCARD', KEYS[1]) < tonumber(ARGV[5])) then
    redis.call('ZADD', KEYS[1], tonumber(ARGV[2]), ARGV[1])
    res = 1
end
redis.call('EXPIRE', KEYS[1], tonumber(ARGV[4]))
return res

ARGV[1]: zset element
ARGV[2]: zset score(當(dāng)前時間戳)
ARGV[3]: 30秒前的時間戳
ARGV[4]: zset key 過期時間30秒
ARGV[5]: 限流閾值

private final RedisTemplate<String, Object> redisTemplate;

public boolean execLuaScript(String luaStr, List<String> keys, List<Object> args){
	RedisScript<Boolean> redisScript = RedisScript.of(luaStr, Boolean.class)
	return redisTemplate.execute(redisScript, keys, args.toArray());
}

測試一下效果

@SpringBootTest
public class ApiApplicationTest {
    @Test
    public void test2() throws InterruptedException{
        String luaStr = "redis.call('ZREMRANGEBYSCORE', KEYS[1], 0, tonumber(ARGV[3]))\n" +
                "local res = 0\n" +
                "if(redis.call('ZCARD', KEYS[1]) < tonumber(ARGV[5])) then\n" +
                "    redis.call('ZADD', KEYS[1], tonumber(ARGV[2]), ARGV[1])\n" +
                "    res = 1\n" +
                "end\n" +
                "redis.call('EXPIRE', KEYS[1], tonumber(ARGV[4]))\n" +
                "return res";
        for (int i = 0; i < 10; i++) {
            boolean res = execLuaScript(luaStr, Arrays.asList("aaaa"), Arrays.asList("ele"+i, System.currentTimeMillis(),System.currentTimeMillis()-30*1000, 30, 3));
            System.out.println(res);
            Thread.sleep(5000);
        }
    }
}

測試結(jié)果符合預(yù)期！

擴展閱讀

lua腳本每次都需要傳一長串腳本內(nèi)容來回傳輸，會增加網(wǎng)絡(luò)流量和延遲，而且每次都需要服務(wù)器重新解釋和編譯，效率較為低下。因此，不建議在實際生產(chǎn)環(huán)境中直接執(zhí)行l(wèi)ua腳本，而應(yīng)該使用lua腳本的hash值來進(jìn)行傳輸。

為了方便使用，我們先把方法封裝一下

import lombok.RequiredArgsConstructor;
import org.springframework.data.redis.connection.RedisScriptingCommands;
import org.springframework.data.redis.connection.ReturnType;
import org.springframework.data.redis.core.RedisTemplate;
import org.springframework.data.redis.serializer.RedisSerializer;
import org.springframework.stereotype.Component;

import java.util.List;

/**
 * @author 敖癸
 * @formatter:on
 * @since 2024/3/25
 */
@Component
@RequiredArgsConstructor
public class RedisService {

    private final RedisTemplate<String, Object> redisTemplate;
    private static RedisScriptingCommands commands;
    private static RedisSerializer keySerializer;
    private static RedisSerializer valSerializer;
    
    public String loadScript(String luaStr) {
        byte[] bytes = RedisSerializer.string().serialize(luaStr);
        return this.getCommands().scriptLoad(bytes);
    }

    public <T> T execLuaHashScript(String hash, Class<T> returnType, List<String> keys, Object[] args) {
        byte[][] keysAndArgs = toByteArray(this.getKeySerializer(), this.getValSerializer(), keys, args);
        return this.getCommands().evalSha(hash, ReturnType.fromJavaType(returnType), keys.size(), keysAndArgs);
    }

    private static byte[][] toByteArray(RedisSerializer keySerializer, RedisSerializer argsSerializer, List<String> keys, Object[] args) {
        final int keySize = keys != null ? keys.size() : 0;
        byte[][] keysAndArgs = new byte[args.length + keySize][];
        int i = 0;
        if (keys != null) {
            for (String key : keys) {
                keysAndArgs[i++] = keySerializer.serialize(key);
            }
        }
        for (Object arg : args) {
            if (arg instanceof byte[]) {
                keysAndArgs[i++] = (byte[]) arg;
            } else {
                keysAndArgs[i++] = argsSerializer.serialize(arg);
            }
        }
        return keysAndArgs;
    }

    private RedisScriptingCommands getCommands() {
        if (commands == null) {
            commands = redisTemplate.getRequiredConnectionFactory().getConnection().scriptingCommands();
        }
        return commands;
    }

    private RedisSerializer getKeySerializer() {
        if (keySerializer == null) {
            keySerializer = redisTemplate.getKeySerializer();
        }
        return keySerializer;
    }

    private RedisSerializer getValSerializer() {
        if (valSerializer == null) {
            valSerializer = redisTemplate.getValueSerializer();
        }
        return valSerializer;
    }
}

測試一下：

@SpringBootTest
@TestInstance(TestInstance.Lifecycle.PER_CLASS)
public class ApiApplicationTest implements ApplicationContextAware {

    private static ApplicationContext context;
    private static RedisService redisService;
    public static String luaHash;

    private final static String LUA_STR = "redis.call('ZREMRANGEBYSCORE', KEYS[1], 0, tonumber(ARGV[3]))\n" +
            "local res = 0\n" +
            "if(redis.call('ZCARD', KEYS[1]) < tonumber(ARGV[5])) then\n" +
            "    redis.call('ZADD', KEYS[1], tonumber(ARGV[2]), ARGV[1])\n" +
            "    res = 1\n" +
            "end\n" +
            "redis.call('EXPIRE', KEYS[1], tonumber(ARGV[4]))\n" +
            "return res";

    @Override
    public void setApplicationContext(ApplicationContext applicationContext) throws BeansException {
        context = applicationContext;
    }

    @BeforeAll
    public static void before(){
        redisService = context.getBean(RedisService.class);
        luaHash = redisService.loadScript(LUA_STR);
        System.out.println("lua腳本hash: "+ luaHash);
    }


    @Test
    public void testLuaHash() throws InterruptedException {
        for (int i = 0; i < 50; i++) {
            List<String> keys = Collections.singletonList("aaaa");
            Object[] args = new Object[]{"ele" + i, System.currentTimeMillis(), System.currentTimeMillis() - 30 * 1000, 30, 3};
            Boolean b = redisService.execLuaHashScript(luaHash, Boolean.class, keys, args);
            System.out.println(b);
            Thread.sleep(3000);
        }
    }
}

使用的時候在項目啟動時候，把腳本load一下，后續(xù)直接用hash值就行了