Merge pull request #49 from marklogic/feature/457-error-handling-docs

DEVEXP-457 Added docs for error handling

Author: rjrudin

docs/reading.md

@@ -128,6 +128,17 @@ spark.readStream \
     .load()
 ```
 
+## Error handling
+
+When reading data with the connector, any error that occurs - whether it is before any data is read or during a request
+to MarkLogic to read data - is immediately thrown to the user and the read operation ends without any result being 
+returned. This is consistent with the design of Spark connectors, where a partition reader is allowed to throw 
+exceptions and is not given any mechanism for reporting an error and continuing to read data. The connector will strive 
+to provide meaningful context when an error occurs to assist with debugging the cause of the error. 
+
+In practice, it is expected that most errors will be a result of a misconfiguration. For example, the connection and 
+authentication options may be incorrect, or the Optic DSL query may have a syntax error. Any errors that cannot be 
+fixed via changes to the options passed to the connector should be reported as new issues to this GitHub repository.
 
 ## Tuning performance
 

docs/writing.md

@@ -121,6 +121,22 @@ connector and into MarkLogic. This will result 100 new JSON documents in the `st
 The ability to stream data into MarkLogic can make Spark an effective tool for obtaining data from a variety of data 
 sources and writing it directly to MarkLogic. 
 
+## Error handling
+
+The connector may throw an error during one of two phases of operation - before it begins to write data to MarkLogic, 
+and during the writing of data to MarkLogic. 
+
+For the first kind of error, the error will be immediately returned to the user and no data will have been written. 
+Such errors are often due to misconfiguration of the connector options and should be fixable. 
+
+For the second kind of error, the error will eventually be returned to the user, usually within seconds of it 
+occurring. The slight delay is due to the asynchronous nature of data being written by the connector. The error will 
+be logged by the connector and the write operation will be aborted. Any batches of documents that were written 
+successfully prior to the error occurring will still exist in the database. 
+
+Similar to errors with reading data, the connector will strive to provide meaningful context when an error occurs to 
+assist with debugging the cause of the error. Any errors that cannot be fixed via changes to the options passed to the 
+connector should be reported as new issues to this GitHub repository.
 
 ## Tuning performance