Max Partition Size Best Practices

Guidelines for getting reliable signal from Max Partition Size checks while keeping the noise (and the maintenance) low.

Size the threshold from real loads, not from a round number

Look at what the container actually loads over a few weeks before picking a value. A threshold set from intuition either never fires or fires every day. Leave enough headroom for legitimate variation, such as a month-end peak or a quiet weekend.

Use the pair to bracket the expected range

Max Partition Size on its own only guards one side. Adding Min Partition Size to the same container brackets the load between a floor and a ceiling, which is usually what "the batch looked normal" really means.

Remember the threshold is per partition

On a file container that loads dozens of files per Scan, a threshold sized for the whole delivery will fire on every single file. Size it for one partition.

Reach for Volumetric when you want a trend, not a limit

A fixed threshold answers "is this partition the right size?". Volumetric answers "did this container change size compared to how it usually behaves?", which needs no number to maintain.

Scope the check with the read strategy, not with a filter or coverage

The rule takes neither a filter clause nor a coverage threshold, so there is no way to narrow the evaluation to part of a partition. What the check sees is exactly what the Scan loaded, which means the read strategy on the container is the only lever over the counted set.

Route the anomalies to the right people

A failure usually points at the system that produced the value, not at the warehouse. Set an Anomaly Assignee from the team that owns that producer, and tag the check so related checks are easy to find.

See Also