ANNOUNCE: ScoutFS archival filesystem code published
Greetings -fsdevel,
Today we at Versity are opening the code to ScoutFS, the clustered file system that we've been developing as part of our larger software stack that supports large scale archives.
The motivation for the project and the architectural decisions that we've made can be found in the white paper that is linked off of
https://www.scoutfs.org/ . We've also set up a
[email protected] development mailing list and have an open Slack channel, both are linked off of the scoutfs.org site.
The README.md in the kernel module github repo at
https://github.com/versity/scoutfs-kmod-dev/ describes the quick steps needed to get a system up and running.
For the expert audience, here's the overview of the project:
- Clustered file system using a shared block device.
- Shared LSM indexing of metadata to encourage concurrent updates.
- Integrated archival interfaces (indexing, "offline" extent tracking).
- Batch locking to reduce the cost of enforcing full POSIX.
- Initial development targets RHEL/CentOS kernels.
- What you'd expect: atomic transactions, metadata checksums, extents.
This code can be considered a rough beta. The large architectural structures are there for review, and what is implemented has been well exercised, but a lot remains to be implemented before we declare the format fixed and submit the code upstream.
We're opening the project early to give the community the opportunity to contribute to the design and implementation.
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Ask me anything :),
- z
Quoted 33 lines · Zach Brown
Greetings -fsdevel,
Today we at Versity are opening the code to ScoutFS, the clustered file system that we've been developing as part of our larger software stack that supports large scale archives.
The motivation for the project and the architectural decisions that we've made can be found in the white paper that is linked off of
https://www.scoutfs.org/ . We've also set up a
[email protected] development mailing list and have an open Slack channel, both are linked off of the scoutfs.org site.The README.md in the kernel module github repo at
https://github.com/versity/scoutfs-kmod-dev/ describes the quick steps needed to get a system up and running.For the expert audience, here's the overview of the project:
- Clustered file system using a shared block device.
- Shared LSM indexing of metadata to encourage concurrent updates.
- Integrated archival interfaces (indexing, "offline" extent tracking).
- Batch locking to reduce the cost of enforcing full POSIX.
- Initial development targets RHEL/CentOS kernels.
- What you'd expect: atomic transactions, metadata checksums, extents.This code can be considered a rough beta. The large architectural structures are there for review, and what is implemented has been well exercised, but a lot remains to be implemented before we declare the format fixed and submit the code upstream.
We're opening the project early to give the community the opportunity to contribute to the design and implementation.
Quoted 7 lines
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Ask me anything :),
Is there a fsck tool for this? :D
--D
Quoted 1 line
- z
Quoted 5 lines · Darrick J. Wong
Quoted 2 lines · Zach Brown
Ask me anything :),
Is there a fsck tool for this? :D
Not yet, it's early days :)
- z
Quoted 7 lines · Zach Brown
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Ask me anything :),
How well does it survive xfstests? ;-)
Quoted 10 lines · Matthew Wilcox
Quoted 7 lines · Zach Brown
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Ask me anything :),
How well does it survive xfstests? ;-)
Pretty well? It certainly doesn't run a lot of the tests because features aren't there yet, but it does well at what it does run :).
Not surprisingly, it's been our test platform as we've been developing. We have a branch that adds some cluster awareness and scoutfs tests. Cleaning up that branch and making the repo public is high on my list.
- z
Zach,
Quoted 7 lines · Zach Brown
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Ask me anything :),
Is encryption planned?
Both storage and transport.
Signature · 2 lines
Thanks, //richard
Quoted 9 lines · Richard Weinberger
Zach,
Quoted 6 lines · Zach Brown
In the coming weeks I'll personally be focusing on some big ticket functional items (deleted inode cleanup in particular), hardening a few recovery cases after crashes, and in general spending all of my will power focusing on that responsible nonsense instead of getting lost in satisfying performance tuning.
Quoted 4 lines
Quoted 1 line
Ask me anything :),
Is encryption planned?
Both storage and transport.
Not specifically planned, no, but I'm receptive to adding it -- particularly to support finer granularity key contexts (not just fs mount wide :/).
- z