GridGain Developers Hub

Semaphore

GridGain’s counting distributed semaphore implementation and behavior is similar to the concept of a well-known java.util.concurrent.Semaphore. As any other semaphore it maintains a set of permits that are taken using acquire() method and released with release() counterpart allowing to restrict access to some logical or physical resource or synchronize execution flow. The only difference is that GridGain’s semaphore empowers you to fulfill these kind of actions not only in boundaries of a single JVM but rather a cluster wide, across many remote nodes.

You can create a distributed semaphore as follows:

Ignite ignite = Ignition.start();

IgniteSemaphore semaphore = ignite.semaphore("semName", // Distributed semaphore name.
        20, // Number of permits.
        true, // Release acquired permits if node, that owned them, left topology.
        true // Create if it doesn't exist.
);

Once the semaphore is created, it can be used concurrently by multiple cluster nodes in order to implement some distributed logic or restrict access to a distributed resource like in the following example:


Ignite ignite = Ignition.start();

IgniteSemaphore semaphore = ignite.semaphore("semName", // Distributed semaphore name.
        20, // Number of permits.
        true, // Release acquired permits if node, that owned them, left topology.
        true // Create if it doesn't exist.
);

// Acquires a permit, blocking until it's available.
semaphore.acquire();

try {
    // Semaphore permit is acquired. Execute a distributed task.
    ignite.compute().run(() -> {
        System.out.println("Executed on:" + ignite.cluster().localNode().id());

        // Additional logic.
    });
} finally {
    // Releases a permit, returning it to the semaphore.
    semaphore.release();
}