In Apache Spark, what triggers a shuffle?

A shuffle in Spark is triggered by wide transformations that require data to be redistributed across partitions. Common examples are groupBy, reduceByKey, join, cogroup, and distinct. These operations need to bring all records with the same key to the same executor, which involves network transfer and disk I/O. Narrow transformations like map or filter do not trigger shuffles because they can be performed locally on each partition.