A core UIMA analysis engine has no error handling. Exceptions are passed back to the caller. Remote services accessed from a core aggregate AE add a timeout option which will generate a timeout exception.
Marshall has a scenario where a scaled out UIMA aggregate AE running as a hadoop mapper has a remote delegate. This delegate is implemented as a UIMA AS service, accessed via a JMS client descriptor. In extremely rare cases the remote service takes too long to finish processing.
In order for the hadoop job to complete successfully, Marshall would like a "continue on failure" option, where after a timeout the error is ignored and the original CAS sent to the remote service is used to continue processing in the core UIMA aggregate. This option is easy to implement with UIMA AS services, because a copy of the outstanding CAS is held by the client.
The syntax would be to add an optional argument to the JMS client descriptor, e.g.
<parameter name="ignore_process_errors" value="true"/>