What would you like to happen?
Prior to release 2.37.0, setting the batchSize indicated that a batch would consist of at least that many records if executed prior to the window duration. After the release of 2.37.0, batchSize instead became the minimum batch size, meaning that if the elements in the given bundle did not exceed the set batchSize, the elements would be executed immediately rather than collected to be executed once the batchSize was reached.
This has caused issues in pipelines where more frequent writes are causing more deadlocks/lock wait timeouts. The problem then gets exacerbated as backlog grows and workers are added.
I'm proposing that we have a way to at least differentiate between these two batching strategies to allow for batches to build up rather than execute with a very small number of elements.
Issue Priority
Priority: 2 (default / most feature requests should be filed as P2)
Issue Components
What would you like to happen?
Prior to release 2.37.0, setting the
batchSizeindicated that a batch would consist of at least that many records if executed prior to the window duration. After the release of 2.37.0,batchSizeinstead became the minimum batch size, meaning that if the elements in the given bundle did not exceed the setbatchSize, the elements would be executed immediately rather than collected to be executed once thebatchSizewas reached.This has caused issues in pipelines where more frequent writes are causing more deadlocks/lock wait timeouts. The problem then gets exacerbated as backlog grows and workers are added.
I'm proposing that we have a way to at least differentiate between these two batching strategies to allow for batches to build up rather than execute with a very small number of elements.
Issue Priority
Priority: 2 (default / most feature requests should be filed as P2)
Issue Components