Product Use Case

As a user of the MongoDB Spark connector, I may want to ignore duplicate key errors, when inserting data.

As Spark tasks with partial execution can fail, the partition be processed by another executor. When this happens duplicate key exceptions occur.

This makes insert only operations fragile and the alternative is to rely on slower upserts.

User Impact

• Adding this feature would improve throughput for insert only workloads __ 

Acceptance Criteria

Implementation Requirements

• Will only work for unordered writes
• The data must include an _id / the fields declared in the idFieldList
• For insert only operations.
• Support batched and streaming writes