Faulted Task from processing does not shut down processor

[jsmithtx]

Using version 2.2.5 of the library, an unhandled exception that escapes from document processing does not shut down the processor. In tracing through the logic, it appears that the PartitionSupervisor does throw the exception (and sets the close reason as ObserverError, line 66-70),
catch (Exception) when (processorTask.IsFaulted)
{
closeReason = ChangeFeedObserverCloseReason.ObserverError;
throw;
}
But the parent PartitionController eats the exception by simply warning that processing fails (line 132-135),
catch (Exception e)
{
Logger.WarnException("partition {0}: processing failed", e, lease.PartitionId);
}
And the caller of this method isn't awaiting the method (line 71)
this.ProcessPartition(supervisor, lease).LogException();

This seems like it might be a bug; for our use of the library, we must have at least once processing and previously we were raising exceptions to "blow the stream", but with the new version of the library, this is no longer happening. What is the recommendation to ensure we stop the change feed when there is an exception? Thanks!

[jsmithtx]

Any update on this issue?

[ealsur]

@jiffypopjr If the Observer throws an error, it would shutdown the Observer, not the Processor. The Processor would eventually spin-up another Observer to continue from the last known checkpoint (Lease Continuation).

[jsmithtx]

Ok, we run in service fabric, so we've used the SF capability to blow the process which causes SF to immediately start a new process (which in theory, picks it up again). However, #131 is blocking this from happening.