How does one properly setup Distributed Processing for Summation or eDiscovery?
There are certain things (e.g., Unitization, SWF on the fly, etc.) that require the presence of an engine on the app server (i.e., the server where WCF is installed). In a single server environment, there will be an engine on the app server. When we add DPEs into an environment we have to make sure we install an engine on the app server. This engine MUST be a DPE. It cannot be a local engine.
If there is no engine on the WCF server, things like unitization will fail.
If there is a local processing engine on the WCF server, communication back to the processing manager from the distributed processing engine(s) will be disrupted (i.e., things will not work). Installing a distributed processing engine on the WCF server, but not adding it to the distributed processing manager (DPM) configuration is the correct way.
This is how things need to be set up.
Distributed Processing done right!