In this post, the author explores whether changing the ZFS record size would affect the speed of PostgreSQL tests as the previous tests showed that ZFS was slower for the standard tests.
By testing with a ZFS having an aligned record size to match the PostgreSQL block size, it was found that aligning the PostgreSQL block size and the ZFS record size mattered a lot.
The results showed that using an aligned record size improved the speed considerably, which means that ZFS today is a valid option to run PostgreSQL workloads.
The overall tests included running standard and simple-update tests with a 8KB and 32KB PostgreSQL block size and ZFS record size.
It was found that using a 32KB block- and recordsize gave better results than using an aligned 8KB block- and recordsize.
The author recommends testing your own workload before making any decisions. Some other factors like shared_buffers, checkpointing etc are also to be adapted.