bzfs v1.12.0 – Fleet‑scale ZFS snapshot replication, safer defaults, and performance boosts

[whoschek]

bzfs is a batteries‑included CLI for reliable ZFS snapshot replication using zfs send/receive (plus snapshot creation, pruning, and monitoring). bzfs_jobrunner is the orchestrator for periodic jobs across a fleet of N source hosts and M destination hosts.

Highlights in 1.12.0:

• Fleet‑scale orchestration: bzfs_jobrunner is now STABLE and can replicate across a fleet of N source hosts and M destination hosts using a single shared job config. Ideal for geo‑replication, multi‑region read replicas, etc.
• Snapshot caching that "just works": --cache-snapshots now boosts replication and --monitor-snapshots.
• Find latest common snapshot even among non‑selected snapshots (more resilient incrementals).
• Better scheduling at scale: new --jitter to stagger starts; per‑host logging; visibility of skipped subjobs; --jobrunner-dryrun; --jobrunner-log-level; SSH port/config options; tighter input validation.
• Bookmark policy made explicit: replace --no-create-bookmarks with --create-bookmarks={none,hourly,minutely,secondly,all} (default: hourly).
• Security & safety:
  • New --preserve-properties to retain selected dst properties across replication.
  • Safer defaults: zfs send no longer includes --props by default; instead a safe whitelist of properties is copied on full sends via zfs receive -o ... options.
  • Prefer --ssh-{src|dst}-config-file for SSH settings; stricter input validation; private lock dirs; tighter helper constraints; refuse symlinks; ssh -v when using -v -v -v.
• Performance and UX:
  • Parallel detection of ZFS features/capabilities on src+dst; parallel bookmark creation.
  • Auto‑disable mbuffer and compression on loopback; improved local‑mode latency.
  • Robust progress parsing for international locales; cleaner shutdown (propagate SIGTERM to descendants).
• Quality of life: bash completion for both bzfs and bzfs_jobrunner; docs and nightly tests updates.

Other notable changes:

• Support --delete-dst-snapshots-except also when the source is not a dummy.
• Log more detailed diagnostics on --monitor-snapshots.
• Run nightly tests also on zfs-2.3.4, zfs-2.2.8 and FreeBSD-14.3

Changes to watch for (deprecations & migration):

• bzfs_jobrunner:
  • --jobid replaced by required --job-id and optional --job-run (old name works for now; will be removed later).
  • --replicate no longer needs an argument (the argument is deprecated and ignored).
  • --src-user / --dst-user renamed to --ssh-src-user / --ssh-dst-user (old names deprecated).
• bzfs:
  • --create-src-snapshots-enable-snapshots-changed-cache replaced by --cache-snapshots.
  • --no-create-bookmarks replaced by --create-bookmarks=… as above.
  • If you relied on zfs send --props by default, re‑enable the old behavior explicitly, for example:
    • --zfs-send-program-opts="--props --raw --compressed" --zfs-recv-o-targets=full+incremental
• Installation via pip remains unchanged. Optional system installation from the git repo is now done by adding symlinks to the startup
  shell scripts.

Install / Upgrade:

pip install -U bzfs

# or run from git without system install:
git clone https://github.com/whoschek/bzfs.git
cd bzfs/bzfs_main
./bzfs --help
./bzfs_jobrunner --help
sudo ln -sf $(pwd)/bzfs /usr/local/bin/bzfs  # Optional system installation
sudo ln -sf $(pwd)/bzfs_jobrunner /usr/local/bin/bzfs_jobrunner  # Optional system installation

Links:

• Detailed Changelog: https://github.com/whoschek/bzfs/blob/main/CHANGELOG.md
• README (bzfs): https://github.com/whoschek/bzfs#readme
• README (bzfs_jobrunner): https://github.com/whoschek/bzfs/blob/main/README_bzfs_jobrunner.md
• PyPI: https://pypi.org/project/bzfs/

As always, please test in a non‑prod environment first. Feedback, bug reports, and ideas welcome!