Which is the recommended record size for a zfs dataset hosting a Netatalk CNID backend database?? (.AppleDB)

The plan is to use a zfs dataset exclusively for the Netatalk CNID backend, and configure the recordsize property according.

CNID backend database is based in Berkeley DB.


  • Benchmark the performance of various ZFS recordsizes using your projected workload, and pick the best one. – Andrew Henle Mar 7 at 10:31
  • 1
    Typically, smaller block sizes are better for performance because they do not cause as much write amplification (updating a sub-block doesn't create a read-modify-write), but they are worse for compression. I would try 4K as a reasonable guess, but I agree there is no clear way to know without benchmarking your actual workload, since it appears BerkeleyDB can be configured to use block sizes from 512B to 64KB. – Dan Mar 7 at 21:12
  • However, you should at least make sure you use the same block size for both BerkeleyDB and your filesystem, and that both of those are larger than the block size your disks are using. – Dan Mar 7 at 21:13

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.