redisson / redisson Goto Github PK

Operates with async methods only. Code example:

RBatch batch = redisson.createBatch();

RMap<Integer, Integer> map = redisson.getMap("map");
map.fastPutAsync(1, 2);
map.putAsync(2, 3);
map.fastPutAsync(3, 4);

List<Future<?>> futures = batch.execute();
// or
Future<List<Future<?>>> futures = batch.executeAsync();

Any plan to add TTL to a Lock operation? Don't confuse with tryLock with TIME.

I refer to a situation where a thread is dead and leave a resource locked (deadlock).

Hi
I could not find a way to set expiry for the contents of the cache from within Java. Is this true, or did I miss something?

Thank you
Jozsef

The eval method does not seem to work for me. I traced the cause to this line in theRedisAsyncConnection:eval() method:

args.addValue(script).add(keys.length).addKeys(keys).addValues(values)

It should instead be:

args.add(script).add(keys.length).addKeys(keys).addValues(values)

In its current form, the script is escaped with double quotes and cannot be interpreted by redis.

It would be nice if the Redisson class exposed its connection manager. The use case is to access low level redis commands using the same pool of connections as the high-level redisson abstractions. Currently its protected and private which means I have to resort to ugly hacks to get to it.

Use hmget

After running for a week or two, the scores collapsed to zeros or nead zero (10^-300) and we encountered an infinite loop in redisson when adding (it couldnt settle on a score). So we will manage the score ourselves.

Hi there!
As I understand, Future returned by RedisAsyncConnection moves API user back to the blocking world. Java futures are NOT non-blocking by nature, they have blocking get(). Thus API user needs to spawn new threads and say hello to threading overhead again. Then why Netty is used?
I propose to return promises (Java's 8 CompletableFuture, Guava's ListenableFuture or jdeferred's promises).

Thanks,
Denis

I have test code for the load balance, but only the first instance got the load:

public static void main(String[] args) {
Config config = new Config();
//config.setConnectionPoolSize(100); // default 100
// Redisson will use load balance connections between listed servers
String[] redisInstances = new String[]{"192.168.9.101:6379", "192.168.9.102:6379"};
config.addAddress(redisInstances);

    Redisson redisson = Redisson.create(config);
    Queue<Integer> q = redisson.getQueue("QUEUE");

    for (int i = 0; i < 1_000_000; i++) {
        q.add(i);
        if (i % 10000 == 0) {
            System.out.println("put: " + i);
        }
    }

    System.out.println("DONE.");
    redisson.shutdown();
}

Do I miss any thing?

Hi Nikita, when I ran some load test, ConnectionWatchdog did not restore PubsubConnection correctly (my env is redis 2.8.3, java 1.7.0_45,Netty 4.0.19 final, OSX 10.9.2). After some digging, I found a couple of problems,

1. When ChannelInitilizor ChannelInitializer.initChannel is called, pipeLine.get(CommandHandler.class) and pipeLine.get(RedisAsyncConnection.class) return null, I guess the old ChannelPipeline had been destroyed.
2. The connection retry and backoff doesn’t seem right, the connection attempt seems only fired when the channel is inactive, if re-connect did not work, no further attempt will be fired.
3. RedissonTopic.promise.get().setSuccess(true) will throw IllegalStateException: complete already when PubsubConnection is re established.
   4)Though this is not a bug, it seems unnecessary to use a dedicated HashedTimeWheel to schedule reconnection task.

I will push a patch later.

Redisson clusters may return a 'MOVED' if a key is on another shard, does redisson handle that?

I'm trying to use Lock and Unlock on Jersey Resource.

1. User makes a POST request
2. System LOCK "A"
3. System does some stuff
4. System UNLOCK "A"
5. User makes another POST request
6. System LOCK "A"
7. System does some stuff
8. System UNLOCK "A"

The system crashes at point 6 (view attachment).

If i try to make some LOCK-UNLOCK in a while loop it works, but when i make these LOCK-UNLOCK from different Threads it does not works.

WARN [2014-07-10 10:36:15,734] io.netty.channel.DefaultChannelPipeline: An exceptionCaught() event was fired, and it reached at the tail of the pipeline. It usually means the last handler in the pipeline did not handle the exception.
! java.lang.IllegalStateException: complete already: DefaultPromise@3c519764(incomplete)
! at io.netty.util.concurrent.DefaultPromise.setSuccess(DefaultPromise.java:406) ~[netty-common-4.0.19.Final.jar:4.0.19.Final]
! at org.redisson.RedissonLock$1.subscribed(RedissonLock.java:177) ~[redisson-1.1.3.jar:na]
! at com.lambdaworks.redis.pubsub.RedisPubSubConnection.channelRead(RedisPubSubConnection.java:132) ~[redisson-1.1.3.jar:na]
! at io.netty.channel.DefaultChannelHandlerContext.invokeChannelRead(DefaultChannelHandlerContext.java:341) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.DefaultChannelHandlerContext.fireChannelRead(DefaultChannelHandlerContext.java:327) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at com.lambdaworks.redis.pubsub.PubSubCommandHandler.decode(PubSubCommandHandler.java:46) [redisson-1.1.3.jar:na]
! at com.lambdaworks.redis.protocol.CommandHandler.channelRead(CommandHandler.java:52) [redisson-1.1.3.jar:na]
! at io.netty.channel.DefaultChannelHandlerContext.invokeChannelRead(DefaultChannelHandlerContext.java:341) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.DefaultChannelHandlerContext.fireChannelRead(DefaultChannelHandlerContext.java:327) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.ChannelInboundHandlerAdapter.channelRead(ChannelInboundHandlerAdapter.java:86) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.DefaultChannelHandlerContext.invokeChannelRead(DefaultChannelHandlerContext.java:341) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.DefaultChannelHandlerContext.fireChannelRead(DefaultChannelHandlerContext.java:327) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.DefaultChannelPipeline.fireChannelRead(DefaultChannelPipeline.java:785) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.nio.AbstractNioByteChannel$NioByteUnsafe.read(AbstractNioByteChannel.java:126) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.nio.NioEventLoop.processSelectedKey(NioEventLoop.java:507) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.nio.NioEventLoop.processSelectedKeysOptimized(NioEventLoop.java:464) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.nio.NioEventLoop.processSelectedKeys(NioEventLoop.java:378) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.channel.nio.NioEventLoop.run(NioEventLoop.java:350) [netty-transport-4.0.19.Final.jar:4.0.19.Final]
! at io.netty.util.concurrent.SingleThreadEventExecutor$2.run(SingleThreadEventExecutor.java:116) [netty-common-4.0.19.Final.jar:4.0.19.Final]
! at java.lang.Thread.run(Thread.java:722) [na:1.7.0_10-ea]

Just an idea, would this be possible? Still new to Redis / Redisson.

redis monitor shows the setex followed by the get

1402351706.438560 [0 107.21.134.1:43779] "SETEX" "\x03\x01api:sessions:dE9K23ZT4de6MHMn2V-mg\xf7" "2591977" "\xae\x02\x01o\x01\x00\xa4\x84\xe7\xaa\xd0Q\x01dE9K23ZT4de6MHMn2V-mg\xf7\xc6\x84\xe7\xaa\xd0Q\x01207.253.217.242,10.226.123.45, 54.87.64.23\xb0\x80\xb4\xbc\x02\x05\xaf\x02\x01\x01okoMskkfJxo9JcOzF7Bpi\xf7"
1402351707.163372 [0 107.21.134.1:43779] "GET" "\x03\x01api:sessions:dE9K23ZT4de6MHMn2V-mg\xf7"

the code:

        RBucket<StoredSession> bucket = redisson.<StoredSession>getBucket(PREFIX + id);
        try {
            return bucket.get();
        } catch (RuntimeException e) {
            LOGGER.log(Level.WARNING, "Removing malformed session " + id + ": " + e.getMessage(), e);
            bucket.delete();
            return null;
        }

code returns null.

Implement with binary insertion algorithm

Hi,
why do you have imported lettuce code into your project instead of referencing it via artifact?

The approach your library works is quite cool providing collections. It's just somehow annoying having no access to the underlying RedisClient. Redisson.create(RedisClient client) or even redisson.getList(RedisClient client, String name) etc. would be a way opener approach. The current way, it's Redisson who manages all the Redis stuff for you. I see Redisson as a sort of valueable add-on.

I do not have Jackson and do not want to use it. Sadly, the Config class has a hard pointer on it:

private RedissonCodec codec = new JsonJacksonCodec();

Thus when creating a new Config instance I have the exception:

Caused by: java.lang.NoSuchFieldError: WRITE_BIGDECIMAL_AS_PLAIN
    at org.redisson.codec.JsonJacksonCodec.createObjectMapper(JsonJacksonCodec.java:90)
    at org.redisson.codec.JsonJacksonCodec.<init>(JsonJacksonCodec.java:43)
    at org.redisson.Config.<init>(Config.java:44)
    at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
    at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:62)
    at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
    at java.lang.reflect.Constructor.newInstance(Constructor.java:408)
    at org.codehaus.groovy.reflection.CachedConstructor.invoke(CachedConstructor.java:77)

Codecs should be made optional or at least the default should be the Serializer codec which only depends on the JDK, or a simple String <=> String codec.

For now I cannot use Redisson properly :-(

Pub a message doesn't require a pubsub connection, however get a RTopic does. There are many cases just want to publish a message, get a RTopic from Redisson will automatically subscribe on the channel, which is unnecessary and there is no safe way to close topic after publish message since there may be some listeners are actually interested in the topic.

BTW, RedisConnection has a method publish(K channel, V message)
Unfortunelty the underlying connection manager is not exposed, see issue #13
Provide access to the underlying connection manager

1.Sentinel not support auth password.（I have solve this problem）
2.When redis sentinel occur failover，redisson can not failover to new master automatic。

in RedisClient.java

try {
        final ConnectionWatchdog watchdog = new ConnectionWatchdog(bootstrap, channels, timer);

        ChannelFuture connect = null;
        // TODO use better concurrent workaround
        synchronized (bootstrap) {
            connect = bootstrap.handler(new ChannelInitializer<Channel>() {
                @Override
                protected void initChannel(Channel ch) throws Exception {
                    ch.pipeline().addLast(watchdog, handler, connection);
                }
            }).connect();
        }
        connect.sync();

        watchdog.setReconnect(true);

        return connection;
    } catch (Throwable e) {
        throw new RedisException("Unable to connect", e);
    }

this "synchronized " protect what?

Hi,
I want to use this inside Android, is it supported?

Thanks

Does reddison support the ability to add both a master and slave addresses and only write to the master?

Same client code worked with 1.4 but not 1.5

It would be really nice if you can add support for PFADD | PFCOUNT | PFMERGE commands in Redisson.

I just added requirepass and masterauth to my configuration for 3 slaves and a master. Using 1.5 I am able to write just fine, but when I tried Map::get I get a RedisException: "NOAUTH Authentication required."

I've tried just using a single connection to the master, a 'useMasterSlaveConnection' with only 'setMasterAddress', and a 'useMasterSlaveConnection' with all hosts configured.

When I connect with a tool like Redis Desktop Manager I am able to view the updated values just fine.

Does redisson support local cache?

For some usage scenario, local read is very heavy, local write is not much, and the delay of data synchronization can be tolerated in seconds. Some configurable cache mechanism could be helpful.

Hi,

When we create a Redisson object, I would also need to execute some Redis commands.

What is the best way ? Do I need to create another ConnectionManager completely separate from the one in the Redisson class, or it would be better to access the one created on the Redisson object, so that we can re-use it ?

Thanks!

I just imported redisson 1.0.3 via maven, set up a redis-server on localhost on default port and wanted to try the redisson redis client. My test code looks as following:

Redisson redis = Redisson.create();
Map m = redis.getMap("someMap");

But I'm getting an Error:

java.lang.NoSuchFieldError: WRITE_BIGDECIMAL_AS_PLAIN
    at org.redisson.codec.JsonJacksonCodec.createObjectMapper(JsonJacksonCodec.java:90)
    at org.redisson.codec.JsonJacksonCodec.<init>(JsonJacksonCodec.java:43)
    at org.redisson.Config.<init>(Config.java:44)
    at org.redisson.Redisson.create(Redisson.java:85)
    at ...

Am I doing something wrong or is this a dependency problem?

In RBucket we have:

    @Override
    public V get() {
        RedisConnection<String, V> conn = connectionManager.connectionReadOp();
        try {
            return conn.get(getName());
        } finally {
            connectionManager.release(conn);
        }
    }

1. get connection
2. do something
3. return back to pool

Before using this, I was simply getting a RedisConnection from the connection manager like this:

        Field field = Redisson.class.getDeclaredField("connectionManager");
        field.setAccessible(true);
        return (ConnectionManager) field.get(redisson);

The RedisConnection class javadoc states that the connection is thread-safe. So I was keeping a singleton reference over one connection to issue some simple commands like get, set, del, ... from several threads.

Questions:

1. Is a RedisConnection class is intended to be used this way (multiple threads doing get/set/del ops) ?
2. If yes, would it be possible to expose this behavior so that users could get from Redisson a connection (with auto-reconnect) ? This connection could be kept as a singleton ref during the entire lifetime of the server and not be returned back to the pool, because this connection will be used often (i.e. session management). This will avoid the burden of the get from pool / return to pool code.

Thanks!

Using Distributed Map, put obj is ok, when get some obj by key, an exception will occur:

Exception in thread "main" com.lambdaworks.redis.RedisException: Command timed out
at com.lambdaworks.redis.RedisAsyncConnection.await(RedisAsyncConnection.java:1179)
at com.lambdaworks.redis.RedisConnection.await(RedisConnection.java:843)
at com.lambdaworks.redis.RedisConnection.hget(RedisConnection.java:283)
at org.redisson.RedissonMap.get(RedissonMap.java:89)
at dispatcher.RedisRef.main(RedisRef.java:573)

the exception will not occur when get naive java obj (String, int, ...)

See GIST: https://gist.github.com/mathieucarbou/21051be4a2abd7cc1006

Note: I don't really know if there is a better place to ask for these sorts of questions... ;-)

Besides all the cool features added to Redisson to use locks, queues, maps, etc... How does Redisson performs compared to Jedis in terms of:

• memory usage
• speed
• number of concurrent connections supported
• use in multi-threading app requiring in both case using a pool
• thread-safety
• etc...

For having used both, I strongly prefer the api of Redisson (using publish subscribe in a reliable way with Jedis is a pain, and also there is no codec). But I am wondering, since their implementation only rely on simple sockets and a pool, how it compares to Redisson, which is based on Netty.

Also concerning thread-safety, having already aked a question about it, and in both libraries a connection pool is used for each operation and it is better to do so. So does thread-safety on a connection level really used ?

Thanks!

It would be great to see TIME supported.

See also #19 in lettuce.