Memory leak when using remoting/actors

[ErikRBergman]

Proposed way to reproduce:

API calling an actor that calls 2-3 other actors and returns data to the API.

Load testing the API should at about 100k requests increase idle CPU usage. Dumping the process shows a a lot of timers.

Details:
Microsoft.ServiceFabric.Actors.Client.ActorProxyFactory is instantiated with with each request through the factory in SetInnerFactory (FG.ServiceFabric.Actors.Client. ActorProxyFactory).

[cedheim]

Reproduction https://github.com/cedheim/FG.ServiceFabric.MemoryLeak
After around 25k calls on a local 1-node cluster the CPU won't drop below 25%

[ErikRBergman]

I've created a fix for this problem. I will need some extensive testing outside my lab environment. Memory stays rather low < 500mb for a full run. A full run took too much time to complete earlier.

The code is pushed to a branch called MemoryLeakFix, https://github.com/FabricatorsGuild/FG.ServiceFabric/tree/MemoryLeakFix based on the latest release.

Who's up for some testing? :)

[ErikRBergman]

Works well with 5 node local cluster as well, and killing of the node processes during testing does not affect the progress very much. So, testing in another project would be suitable.

[ErikRBergman]

Thanks for the repro repo, btw :)