Hazelcast Ringbuffer readManyAsync returns Empty Results

Question

I'm trying to read N items from a RingBuffer using readManyAsync but It's always returns an empty resultSet. If I use readOne I get data.

I'm using the readManyAsync as the documentation specify. There is another way to do that?

Enviroment:

• Java 8
• Hazelcast 3.5.3

Example:

Ringbuffer<String> buffer = this.hazelcastInstance.getRingbuffer("testBuffer");
buffer.add("a");
buffer.add("b");
buffer.add("c");
Long sequence = buffer.headSequence();
ICompletableFuture<ReadResultSet<String>> resultSetFuture = buffer.readManyAsync(sequence, 0, 3, null);
ReadResultSet<String> resultSet = resultSetFuture.get();

System.out.println("*** readManyAsync *** readCount: " + resultSet.readCount());
int count = 0;
for (String s : resultSet) {
    System.out.println(count + " - " + s);
    count++;
}

System.out.println("*** readOne ***");
for (int i = 0; i < 3; i++) {
    System.out.println(i + " - " + buffer.readOne(i));
}

Output:

*** readManyAsync *** readCount: 0
*** readOne ***
0 - a
1 - b
2 - c

Answer 1

You are happy with receiving zero results:

buffer.readManyAsync(sequence, 0, 3, null);

Try changing 0 to 1.

buffer.readManyAsync(sequence, 1, 3, null);

Now the call will block till there is at least 1 result.

Probably you can make things more efficient by asking for more than 3 items. In most cases, retrieving data is cheap, but the io/operation scheduling is expensive. So try to batch as much as possible. So try to get as many results as possible.. e.g. 100... or 1000 (which is the max).

Answer 2

Ok, but how do you use readManyAsync in a non-blocking way, with minCount to 0 ?

I made a minimal test case, and I really can't figure it out. I posted a support topic here :

https://groups.google.com/forum/#!topic/hazelcast/FGnLWDGrzb8

As an answer : I use readManyAsync with a timeout, like so :

try{
  buffer.readManyAsync(sequence, 1, 3, null).get(500, TimeUnit.MILLISECONDS);
} catch (TimeoutException e){
  // We timed out, shame, let's move on
}

That seems the only way to make a graceful non-blocking thread, but by reading the doc, I really thought a minCount=0 would do the trick.