Package: wnpp Severity: wishlist Owner: Urmas Rist <ur...@urist.ee> X-Debbugs-Cc: debian-de...@lists.debian.org
* Package name : saunafs Version : 4.4.0 Upstream Contact: Name <supp...@saunafs.com> * URL : https://github.com/leil-io/saunafs * License : GPL Programming Lang: C, C++ Description : SaunaFS is a distributed, FUSE filesystem that is focused on reliability and data integrity. Some of the features include: * Erasure-coding (EC) for files. * Real-time journaling for recovery of metadata and auditing. * Copy-on write snapshots of files. * Extra protection when reading/writing data through CRC. * High-availability and optional automatic failover to one or more shadow master servers. Disclaimer: I am the upstream maintainer and employed to work on it. My employer was asked by some in the Debian Medical mailing list (https://lists.debian.org/debian-med/2024/07/msg00030.html) to help package this for Debian. While the company declined (since we don't have the manpower for it), I decided I would do help out in my personal free time and capacity. SaunaFS was forked from LizardFS (which was originally forked from MooseFS) a year and half ago, when development completely stopped (in reality, development was already slow to non-existant years prior, or it was closed-source). The developers (most of who worked on LizardFS during the last days of it, and some who worked in the very first days) are employed by the company Leil Storage, which is selling storage hyperscaler solutions. Since forking from LizardFS, there have been many fixes and features added to SaunaFS, and is currently more maintained than LizardFS was in the last 4 years. However, it doesn't support upgrading from LizardFS to SaunaFS directly currently (there are plans to add some level of support for migrating from LizardFS, like metadata conversion). It is similar to MooseFS (they share a lot of the codebase from MooseFS 1.6), however it has a few key advanatages in terms of culture: 1. Many developers have worked on it over the years (from LizardFS to SaunaFS), and has such the code is much more maintainable than MooseFS, which only one developer has worked on. 2. All of the tests (around 200-300) for SaunaFS/LizardFS is open source, whereas MooseFS has four tests and unknown (if any) tests that are closed-sourced, making working on the code difficult in a open-source environment. 3. Combined with the above factors, any fork of MooseFS would be difficult. 4. Many features (e.g EC) and new releases of MooseFS are only offered under proprietary licenses. SaunaFS meanwhile releases frequently (generally around once a month) completely free and open source, with some proprietary plugins for things like SMR drives outside the project. 5. MooseFS project health is also concerning, given that only one developer works on it and if he stops working, other developers would have a difficult time working on it. Thus, SaunaFS could be viewed as a more open-source and free alternative to MooseFS that is being actively maintained and will continue to be for the likely forseeable future. I would start packaging by reviewing the previous LizardFS packages. Some things are likely not relevant anymore. For example, renaming the binaries to not conflict with MooseFS (this was done upstream). I plan to spend about 16 hours per month helping maintain this package. No plans are currently do it in a packaging under a team, however Debian med could be a candidate since this ITP was asked on there. I'm also looking for a sponsor. This package would probably split up into their own sub packages (e.g saunafs-master, saunafs-chunkserver) similar to MooseFS/LizardFS.
