An (almost) catastrophic OpenZFS bug and the humans that made it (and Rust is here too) • Rob Norris
https://despairlabs.com/blog/posts/2025-07-10-an-openzfs-bug-and-the-humans-that-made-it/
An almost catastrophic OpenZFS bug and the humans that made it
https://despairlabs.com/blog/posts/2025-07-10-an-openzfs-bug-and-the-humans-that-made-it/
A few weeks ago I fixed a bug in #OpenZFS and I can't stop thinking about it.
https://despairlabs.com/blog/posts/2025-07-10-an-openzfs-bug-and-the-humans-that-made-it/
The recording of the July 9th, 2025 #OpenZFS Production User Call is up:
We discussed pam_zfs_keys and encrypted home directories, a recent FreeBSD multipathing discussion, configuration of a new system, DRAID, arbitrary RAIDZ, and more!
"Don't forget to slam those Like and Subscribe buttons."
Kubuntu shutting down gracefully – without forcing off the computer – following an insane zpool-scrub(8) command:
https://github.com/openzfs/zfs/issues/14481#issuecomment-3053530299
I switched to Kubuntu with root-on-ZFS:
― https://www.reddit.com/r/freebsd/comments/1lr4ayi/switched/
FreeBSD is not entirely abandoned:
― https://www.reddit.com/r/freebsd/comments/1lr4ayi/comment/n1cnfy2/?context=1
Pinned (seeking guidance):
― https://mastodon.bsd.cafe/@grahamperrin/114640816892663471
Added 𝗨𝗣𝗗𝗔𝗧𝗘 𝟮 - 𝗜𝗻𝘁𝗲𝗿𝗶𝗺 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻 [UPDATE 2 - Interim Solution] to 𝗙𝗮𝗶𝗹𝗲𝗱 𝗕𝗮𝗰𝗸𝘂𝗽 𝗦𝗲𝗿𝘃𝗲𝗿 𝗕𝘂𝗶𝗹𝗱 [Failed Backup Server Build] article.
https://vermaden.wordpress.com/2025/05/28/failed-backup-server-build/#interim
Added 𝗨𝗣𝗗𝗔𝗧𝗘 𝟮 - 𝗜𝗻𝘁𝗲𝗿𝗶𝗺 𝗦𝗼𝗹𝘂𝘁𝗶𝗼𝗻 [UPDATE 2 - Interim Solution] to 𝗙𝗮𝗶𝗹𝗲𝗱 𝗕𝗮𝗰𝗸𝘂𝗽 𝗦𝗲𝗿𝘃𝗲𝗿 𝗕𝘂𝗶𝗹𝗱 [Failed Backup Server Build] article.
https://vermaden.wordpress.com/2025/05/28/failed-backup-server-build/#interim
The recording of the July 2nd, 2025 #OpenZFS Production User Call is up:
We discussed desired channel program features, an update to Gibran's VM configuration experiments, a suggestion that bectl allow for % range destroy syntax, and more!
"Don't forget to slam those Like and Subscribe buttons."
Well that was an unexpected way to spend nine hours.
![[Sponsors: Klara, Inc., Wasabi Technology, Inc.]
Motivation and Context
Since 246e588, gang blocks written to raidz vdevs will write past the end of their allocation, corrupting themselves, other d...](https://media.mastodontech.de/cache/preview_cards/images/020/313/241/original/98eb8cd7ee59a3e7.png "[Sponsors: Klara, Inc., Wasabi Technology, Inc.]
Motivation and Context
Since 246e588, gang blocks written to raidz vdevs will write past the end of their allocation, corrupting themselves, other d...")
Hot on the heels of last week's 2.2 update, here's #OpenZFS 2.3.3. Lots of fixes and updates, including:
- encryption+replication panic fixes
- support for Linux 6.15 and RHEL/CentOS/Alma 10
- multithreaded ARC eviction
- more responsive dbuf cache eviction
- improved dedup efficiency for embedded and zero blocks
- syncfs() (Linux) and msync() (FreeBSD) correctness fixes
- disk/by-vdev symlinks for multipath partitions
Enjoy, and remember to tip your programmers! 
The recording of the June 18th, 2025 #OpenZFS Production User Call is up:
We discussed the OpenZFS User and Developer Summit and brainstormed topics for it including libzfs_core, desktop integration, auditing, and more!
"Don't forget to slam those Like and Subscribe buttons."
yaiy, delayed to the battlemesh and now during the train ride getting this on every boot, unless I add a "module_blacklist=zfs"... luckily I'm only using the #ZFS partition for Linux / OpenWrt / Gluon dev stuff, things where I need a deduplicating and compressing filesystem due to many parallel build / git worktrees. And I have a recent backup.
#OpenZFS LTS update. Fairly modest changes, as you would expect from the long-term stable series.
Changes of note:
- all the recent fixes for the infamous encryption+replication panics;
- the "new" Linux IO submission method introduced in 2.3 is now enabled by default, as it fixes a wide variety of issues and is generally faster;
- Linux support up to 6.15.
This is a very pleasant update! If you get your OpenZFS from your OS vendor, hassle them for an update today!
Registration is now open for the 2025 #OpenZFS User and Developer Summit!
@dvl I found #OpenZFS bookmarks *fiendishly* difficult to wrap my head around for a long time, so here's my best attempt at a short and sweet explanation:
A bookmark, unlike a snapshot, doesn't preserve all blocks active at a certain point in time: it merely notes the actual point in time.
You can use a bookmark as a replication SOURCE, and as long as the TARGET has a snapshot taken at the exact time of the bookmark, that is sufficient basis for incremental replication.
At first glance, ZFS snapshots, bookmarks, and checkpoints may seem similar. Let's explore what these features are, how they differ, and example use cases for each:
https://avidandrew.com/zfs-snapshots-bookmarks-checkpoints.html
