add logging for bulk insert and quiet mode to avoid printing tracebacks

[linkdd]

There was no exception catching for bulk insert operations, which would end up with mongo-connector crashing and stopping.

With this feature, I'm adding logging of erronous document's primary key and namespace.

I also added an option quiet (default: False) which, when enable, avoid printing the exception traceback (the feature in pull request #13 can create big requests which is noise in the log file).

[coveralls]

Coverage decreased (-2.7%) to 93.587% when pulling 8257878 on linkdd:feature-log-errors into cacc0f6 on Hopwork:master.

[MaximeGaudin]

Hi @linkdd !

I will not merge this contribution for the following reasons :

• Crashing is easily monitorable (that's what we are doing as a matter of fact) whereas scanning log requires I/O for reading and CPU for pattern matching. Also, it introduces a dependency between the software (via the format of the log) and the monitoring tools (via the pattern to find). A viable solution to this is implementing a health check though
• Crashing was a clear behavior and "guaranteed" to left the base in an incomplete but consistent state. By not crashing, the bulk insert is aborted but the real-time operations are steal streamed to the base. I might be wrong but it seems that you can find yourself with a partial bulk, a gap and the new data.

What do you think ?