Primary Key(s)
An UPSERT behavior is used for the data pushed to Celonis Platform. If the target Celonis Platform table is not defined by the time the connector is created, its schema is determined by the first set of the data file(s) uploaded. If the connector drives the schema, it is recommended to set the primary key configuration entry: connect.ems.data.primary.key. When this is left out, then all the fields are used as part of a composite primary key. Depending on the scenario, this might not be desired.
Apache Kafka guarantees message order at the partition level only. Best practices require the records with the same primary key to be written to one partition.