HashCodeMessageExtractor can create inconsistent ShardId's

[IanPattersonMuo]

Akka.Net 1.3.5

I have a question around the HashCodeMessageExtractor, it uses a GetHashCode on the EntityId which is a string. The problem is that this would appear lead to inconsistent ShardIds between runs of the code. So for example if I create an envelope with an EntityId of 1, the HashCodeMessageExtractor produce a ShardId of 10 consistently for the lifetime of run of the code. However restarting the same code will produce a different ShardId for the same EntityId. When using multiple nodes this would break as the ShardIds for a given EntityId would not be consistent across all of the nodes. This appears to be caused by using the GetHashCode on the EntityId as a String as they not guaranteed to be consistent outside of the app domain.

https://msdn.microsoft.com/en-us/library/system.string.gethashcode(v=vs.110).aspx

The hash code itself is not guaranteed to be stable. Hash codes for identical strings can differ across versions of the .NET Framework and across platforms (such as 32-bit and 64-bit) for a single version of the .NET Framework. In some cases, they can even differ by application domain.

As a result, hash codes should never be used outside of the application domain in which they were created, they should never be used as key fields in a collection, and they should never be persisted.

A simple unit test where the output of the ShardId is verified for consistency shows the problem.

            var extractor = new MessageExtractor();
            var env = new ShardEnvelope(1, "test");
            Assert.Equal("25", extractor.ShardId(env));    

This may work the first time, but subsequent runs of the test fail.

[Horusiath]

@IanPattersonMuo thanks for the info. This indeed may be an issue. We've already met this problem (just in different pieces of project). The fix is quite fast and straightforward - in the following line we should replace GetHashCode with consistent hashing:

akka.net/src/contrib/cluster/Akka.Cluster.Sharding/ClusterSharding.cs

Line 97 in 3f924fd

return (Math.Abs(id.GetHashCode()) % MaxNumberOfShards).ToString();

Inside akka.net there's a MurmurHash.StringHash method, which can be used to generate consistent hashes for strings - we already use them in other parts of the project.