Simple IAffinityFunction implementation leads to crash

I'm using Ignite.NET 2.7.6. I did a very simple implementation of IAffinityFunction interface:

sealed class AffinityFunction : IAffinityFunction
{
  public int Partitions { get { return 1; } }
  public int GetPartition(object key) { return 0; }
  public void RemoveNode(Guid nodeId) {}

  public IEnumerable<IEnumerable<IClusterNode> > AssignPartitions(AffinityFunctionContext context)
  {
    List<IClusterNode> servers = new List<IClusterNode>();
    foreach(IClusterNode node in context.CurrentTopologySnapshot)
    {
      if(!node.IsClient)
      {
        servers.Add(node);
      }
    }

    m_Partitions = new List<IEnumerable<IClusterNode> >();
    if(servers.Count != 0)
    {
      m_Partitions.Add(servers);
    }

    return m_Partitions;
  }

  private List<IEnumerable<IClusterNode> > m_Partitions;
}

Then I assign it to my cache's CacheConfiguration. I must mention that the cache is in the Replicated mode.

Firstly I start server node, then the client node and this leads to server crash with the following error in its log file:

java.lang.IndexOutOfBoundsException: index 0
  at java.util.concurrent.atomic.AtomicReferenceArray.checkedByteOffset(Unknown Source)
  at java.util.concurrent.atomic.AtomicReferenceArray.get(Unknown Source)
  at org.apache.ignite.internal.processors.cache.distributed.dht.topology.GridDhtPartitionTopologyImpl.localPartition0(GridDhtPartitionTopologyImpl.java:911)
  at org.apache.ignite.internal.processors.cache.distributed.dht.topology.GridDhtPartitionTopologyImpl.localPartition(GridDhtPartitionTopologyImpl.java:825)

As I understand this AffinityFunction must collect all the servers in one partition and returns this partition as the one to search in.

Solution

public int Partitions { get { return (m_Partitions != null ? m_Partitions.Count : 0); } }

0 is not a valid value for Partitions property (should be positive)
Partitions property should return a constant value not based on any condition. This property is evaluated once per cache per node.

The error message and documentation should be better, I'll file a ticket.