Like smartd, Debian installs btrfsmaintenance in a disabled state. One day, when consensus on the most useful and safe default configuration has been reached, maintenance operations will be automatically enabled. It also seems wise to wait until btrfsmaintenance supports something like upower, so that these IO and CPU intensive jobs can be paused if a system switches to battery, or deferred until AC power is available. At present btrfsmaintenance is not power-state aware on either laptops or servers with uninterruptible power supplies. Configuration is found at /etc/default/btrfsmaintenance This configuration can be reloaded and its maintenance operations rescheduled by running btrfsmaintenance-refresh.service every time a systemd-enabled system is booted in multi-user mode. To manually reload configuration after editing it run: systemctl restart btrfsmaintenance-refresh.service To refresh maintenance schedule automatically whenever the system is booted (see below for one caveat) run: systemctl enable btrfsmaintenance-refresh.service Enabling that service may also enable btrfsmaintenance-refresh.watch, particularly when enabled by a normal user who then authenticates for org.freedesktop.systemd1.manage-unit-files (root). Refresh.path watches /etc/default/btrfsmaintenance and applies scheduling changes whenever this file is modified. To disable this automatic behaviour run: systemctl disable btrfsmaintenance-refresh.watch Finally, enable systemd timers: systemctl enable btrfs-balance.timer # [1] systemctl enable btrfs-defrag.timer # [2] systemctl enable btrfs-scrub.timer # [3] systemctl enable btrfs-trim.timer # [4] [1] This rewrites all data that is not excluded by a filter. Consensus has not yet been reached on how much nor how often to run balance, and the ideal policy is workload-dependent. Some advocate not running it at all. Eg: Qu Wenruo, writes: Full balance makes no sense. For older kernel we used to recommend balance based on usage filter to free empty block groups, but now empty block groups are automatically freed thus should cause no problem. Balance should be triggered when: - There is very unbalanced data/metadata block groups usage. E.g. a lot of data block groups are almost empty while metadata block groups are almost full. However this should be really rare. - You want to convert profile (https://lore.kernel.org/linux-btrfs/fddf59e6-b4c5-307b-2cb4-fbb8e120ac61@gmx.com/) [2] See wiki.debian.org/Btrfs for issues with defrag eg: out of space errors, complications when using snapshots, fewer more fine-grained paths are better, etc. This is also workload-dependent. Directories with logs, databases, and VM images benefit the most. [3] Recommended; however, battery life will be reduced during a scrub. [4] Prefer fstrim.timer to this option and leave btrfs-trim.timer disabled, in case fstrim.timer is someday enabled by default for non-rotational disks. For the time being, users who forgo systemd can enable roughly equivalent behaviour by running the following script directly: /usr/share/btrfsmaintenance/btrfsmaintenance-refresh-cron.sh To enable this on every boot, add it to /etc/rc.local. To refresh configuration every hour, run: cd /etc/cron.hourly ln -s ../../usr/share/btrfsmaintenance/btrfsmaintenance-refresh-cron.sh ./ Btrfsmaintenance must not be configured to use both the systemd and cron methods (see upstream README). One day, when systemd units for btrfsmaintenance are automatically activated, those who choose not to use systemd will still need to manually activate the refresh-cron script, because upstream maintains that providing automatic scheduling with cron conflicts with automatically enabled scheduling with systemd. -- Nicholas D Steeves , Tue, 4 Aug 2020 10:43:20 -0400