Implement allocation size ranges and use for gang leaves#17111
Merged
Conversation
amotin
reviewed
Mar 3, 2025
amotin
added
the
Status: Revision Needed
github-actions
Bot
removed
the
Status: Revision Needed
amotin
reviewed
Mar 10, 2025
pcd1193182
force-pushed
the
leaf_size
branch
2 times, most recently
from
6a2a041 to
91c6660
Compare
amotin
reviewed
Mar 28, 2025
13 tasks
amotin
reviewed
Apr 23, 2025
When forced to resort to ganging, ZFS currently allocates three child blocks, each one third of the size of the original. This is true regardless of whether larger allocations could be made, which would allow us to have fewer gang leaves. This improves performance when fragmentation is high enough to require ganging, but not so high that all the free ranges are only just big enough to hold a third of the recordsize. This is also useful for improving the behavior of a future change to allow larger gang headers. We add the ability for the allocation codepath to allocate a range of sizes instead of a single fixed size. We then use this to pre-allocate the DVAs for the gang children. If those allocations fail, we fall back to the normal write path, which will likely re-gang. Signed-off-by: Paul Dagnelie <paul.dagnelie@klarasystems.com>
amotin
approved these changes
Apr 23, 2025
tonyhutter
approved these changes
Apr 29, 2025
tonyhutter
added
Status: Accepted
13 tasks
spauka
pushed a commit
to spauka/zfs
that referenced
this pull request
Aug 30, 2025
When forced to resort to ganging, ZFS currently allocates three child blocks, each one third of the size of the original. This is true regardless of whether larger allocations could be made, which would allow us to have fewer gang leaves. This improves performance when fragmentation is high enough to require ganging, but not so high that all the free ranges are only just big enough to hold a third of the recordsize. This is also useful for improving the behavior of a future change to allow larger gang headers. We add the ability for the allocation codepath to allocate a range of sizes instead of a single fixed size. We then use this to pre-allocate the DVAs for the gang children. If those allocations fail, we fall back to the normal write path, which will likely re-gang. Signed-off-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Co-authored-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Tony Hutter <hutter2@llnl.gov>
pcd1193182
added a commit
to KlaraSystems/zfs
that referenced
this pull request
Nov 13, 2025
When forced to resort to ganging, ZFS currently allocates three child blocks, each one third of the size of the original. This is true regardless of whether larger allocations could be made, which would allow us to have fewer gang leaves. This improves performance when fragmentation is high enough to require ganging, but not so high that all the free ranges are only just big enough to hold a third of the recordsize. This is also useful for improving the behavior of a future change to allow larger gang headers. We add the ability for the allocation codepath to allocate a range of sizes instead of a single fixed size. We then use this to pre-allocate the DVAs for the gang children. If those allocations fail, we fall back to the normal write path, which will likely re-gang. Signed-off-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Co-authored-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Tony Hutter <hutter2@llnl.gov>
lundman
pushed a commit
to openzfsonosx/openzfs-fork
that referenced
this pull request
Jan 27, 2026
When forced to resort to ganging, ZFS currently allocates three child blocks, each one third of the size of the original. This is true regardless of whether larger allocations could be made, which would allow us to have fewer gang leaves. This improves performance when fragmentation is high enough to require ganging, but not so high that all the free ranges are only just big enough to hold a third of the recordsize. This is also useful for improving the behavior of a future change to allow larger gang headers. We add the ability for the allocation codepath to allocate a range of sizes instead of a single fixed size. We then use this to pre-allocate the DVAs for the gang children. If those allocations fail, we fall back to the normal write path, which will likely re-gang. Signed-off-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Co-authored-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Tony Hutter <hutter2@llnl.gov>
lundman
pushed a commit
to openzfsonwindows/openzfs
that referenced
this pull request
Feb 16, 2026
When forced to resort to ganging, ZFS currently allocates three child blocks, each one third of the size of the original. This is true regardless of whether larger allocations could be made, which would allow us to have fewer gang leaves. This improves performance when fragmentation is high enough to require ganging, but not so high that all the free ranges are only just big enough to hold a third of the recordsize. This is also useful for improving the behavior of a future change to allow larger gang headers. We add the ability for the allocation codepath to allocate a range of sizes instead of a single fixed size. We then use this to pre-allocate the DVAs for the gang children. If those allocations fail, we fall back to the normal write path, which will likely re-gang. Signed-off-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Co-authored-by: Paul Dagnelie <paul.dagnelie@klarasystems.com> Reviewed-by: Alexander Motin <mav@FreeBSD.org> Reviewed-by: Tony Hutter <hutter2@llnl.gov>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Motivation and Context
This PR pulls out the dynamic allocation size changes from #17004 into their own PR. It also removes the de-ganging logic, which isn't necessary and is difficult to implement without some pretty unpleasant-looking code.
Description
We teach the metaslab code how to allocate a range of sizes, not just a single target size, and use that in the gang block code to try to get it to allocate fewer leaves when possible.
How Has This Been Tested?
Manual testing + the new gang block test.
Types of changes
Checklist:
Signed-off-by.