CODEWITHSUNDEEP
javac++cachingcross-platformignite
Sumit
Sumit

Reputation: 113

Ignite C++ and Cache Affinity

I am using Apache Ignite 1.8.0 for caching on a cluster. I am using the C++ API and am accessing the same cache from both Java and C++. This works fine but I would like to also use affinity collocation to execute tasks on the same node that has cached the data. I am creating the cache in Java, putting the data in C++ but then want to run tasks in Java on this data. The C++ API doesn't have Affinity support so I am now wondering what my options are.

This is how I am creating the cache:

final IgniteCache<Integer, ByteArray> cache = ignite.createCache("myCacheBinaryCpp")

Then I put data from C++. I have a simple byte array class for testing purposes.

int8_t* byteArr= new int8_t[3];
byteArr[0] = 0;
byteArr[1] = 2;
byteArr[2] = 2;
cacheCppJ.Put(i, ByteArray(3,byteArr));

Now I would like to do something like the following but don't know how to ensure that my Java tasks will execute locally to the data.

final Integer affKey = new Integer(9);
ignite.compute().affinityRun("myCacheBinaryCpp", affKey, () -> {
      cache.get(affKey);
      System.out.println("Got cache with affinity");
});

The issue is that the key in C++ is just an int and there is no associated AffinityKey. Therefore I don't know if the affKey I created in Java will always run with the correct node affinity.

Is this the correct approach? I have also considered restricting each of my caches to a pair of nodes to ensure that all operations are on the correct node at least 50% of the time (acceptable me for my use case).

Upvotes: 5

Views: 294

Answers (1)

isapego
isapego

Reputation: 463

If your values serialized properly on the C++ side and can be reached from the Java (you can check it with Put on C++ and Get on Java) then affinityRun will do exactly what it should - it will run task on the primary node for the provided key.

So the answer is "yes, this is going to work".

Upvotes: 1

Related Questions