Add tunable to allow changing micro ZAP's max size

[0mp]

This change turns MZAP_MAX_BLKSZ into a ZFS_MODULE_PARAM() called zap_micro_max_size. As a result, we can experiment with different micro ZAP sizes to improve directory size scaling.

Co-authored-by: Mateusz Piotrowski [email protected]
Signed-off-by: Mateusz Piotrowski [email protected]
Sponsored-by: Wasabi Technology, Inc.

Motivation and Context

Having those additional tunables enables users to experiment with different micro ZAP sizes to improve directory size scaling.

Description

The patch introduces a variable in place of the MZAP_MAX_BLKSZ constant. Additionally, the new variable can be set as a ZFS module parameter (via ZFS_MODULE_PARAM).

How Has This Been Tested?

We've run this patch during some of Klara's recent benchmarks.

gmake cstyle is OK.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Performance enhancement (non-breaking change which improves efficiency)
• Code cleanup (non-breaking change which makes code smaller or more readable)
• Breaking change (fix or feature that would cause existing functionality to change)
• Library ABI change (libzfs, libzfs_core, libnvpair, libuutil and libzfsbootenv)
• Documentation (a change to man pages or other documentation)

Checklist:

• My code follows the OpenZFS code style requirements.
• I have updated the documentation accordingly.
• I have read the contributing document.
• I have added tests to cover my changes.
• I have run the ZFS Test Suite with this change applied.
• All commit messages are properly formatted and contain Signed-off-by.

[0mp]

I forced push to fix the author of the commit (it should be @tsoome first of all) and to resolve a merge conflict in man/man4/zfs.4.

[amotin]

For sizes >= 512KB it may worth to investigate increase the last argument of zfs_btree_create_custom() a bit to avoid additional level in the B-tree. It is a balance between memory usage for small directories, insert/remove cost and search cost.

And I think shrinking of the micro ZAP should be quite easy to implement and good for performance. The bigger you set the maximum size, the more you start wasting memory and CPU without ability to shrink it back.

[0mp]

Rebased and fixed a checkstyle issue (./module/zfs/zap_micro.c: 1730: comma or semicolon preceded by blank).

[0mp]

Thanks!