DOKK / manpages / debian 11 / zfsutils-linux / zfs-mount-generator.8.en
ZFS-MOUNT-GENERATOR(8) System Manager's Manual ZFS-MOUNT-GENERATOR(8)

zfs-mount-generator - generates systemd mount units for ZFS

/lib/systemd/system-generators/zfs-mount-generator

zfs-mount-generator implements the Generators Specification of systemd(1), and is called during early boot to generate systemd.mount(5) units for automatically mounted datasets. Mount ordering and dependencies are created for all tracked pools (see below).

If the dataset is an encryption root, a service that loads the associated key (either from file or through a systemd-ask-password(1) prompt) will be created. This service RequiresMountsFor the path of the key (if file-based) and also copies the mount unit's After, Before and Requires. All mount units of encrypted datasets add the key-load service for their encryption root to their Wants and After. The service will not be Wanted or Required by local-fs.target directly, and so will only be started manually or as a dependency of a started mount unit.

mount unit's Before -> key-load service (if any) -> mount unit -> mount unit's After

It is worth nothing that when a mount unit is activated, it activates all available mount units for parent paths to its mountpoint, i.e. activating the mount unit for /tmp/foo/1/2/3 automatically activates all available mount units for /tmp, /tmp/foo, /tmp/foo/1, and /tmp/foo/1/2. This is true for any combination of mount units from any sources, not just ZFS.

Because ZFS pools may not be available very early in the boot process, information on ZFS mountpoints must be stored separately. The output of the command

zfs list -H -o name,mountpoint,canmount,atime,relatime,devices,exec,readonly,setuid,nbmand,encroot,keylocation,org.openzfs.systemd:requires,org.openzfs.systemd:requires-mounts-for,org.openzfs.systemd:before,org.openzfs.systemd:after,org.openzfs.systemd:wanted-by,org.openzfs.systemd:required-by,org.openzfs.systemd:nofail,org.openzfs.systemd:ignore

for datasets that should be mounted by systemd, should be kept separate from the pool, at

/etc/zfs/zfs-list.cache/POOLNAME

The cache file, if writeable, will be kept synchronized with the pool state by the ZEDLET

history_event-zfs-list-cacher.sh .

The behavior of the generator script can be influenced by the following dataset properties:

If a dataset has mountpoint set and canmount is not off, a mount unit will be generated. Additionally, if canmount is on, local-fs.target will gain a dependency on the mount unit.

This behavior is equal to the auto and noauto legacy mount options, see systemd.mount(5).

Encryption roots always generate a key-load service, even for canmount=off.

Space-separated list of mountpoints to require to be mounted for this mount unit
The mount unit and associated key-load service will be ordered before this space-separated list of units.
The mount unit and associated key-load service will be ordered after this space-separated list of units.
Space-separated list of units that will gain a Wants dependency on this mount unit. Setting this property implies noauto.
Space-separated list of units that will gain a Requires dependency on this mount unit. Setting this property implies noauto.
Toggles between a Wants and Requires type of dependency between the mount unit and local-fs.target, if noauto isn't set or implied.

on: Mount will be WantedBy local-fs.target

off: Mount will be Before and RequiredBy local-fs.target

unset: Mount will be Before and WantedBy local-fs.target

If set to on, do not generate a mount unit for this dataset.


See also systemd.mount(5)

To begin, enable tracking for the pool:

touch /etc/zfs/zfs-list.cache/POOLNAME

Then, enable the tracking ZEDLET:

ln -s "/usr/lib/zfs-linux/zed.d/history_event-zfs-list-cacher.sh" "/etc/zfs/zed.d"

systemctl enable zfs-zed.service

systemctl restart zfs-zed.service

Force the running of the ZEDLET by setting a monitored property, e.g. canmount, for at least one dataset in the pool:

zfs set canmount=on DATASET

This forces an update to the stale cache file.

To test the generator output, run

/lib/systemd/system-generators/zfs-mount-generator /tmp/zfs-mount-generator . .

This will generate units and dependencies in /tmp/zfs-mount-generator for you to inspect them. The second and third argument are ignored.

If you're satisfied with the generated units, instruct systemd to re-run all generators:

systemctl daemon-reload

zfs(5) zfs-events(5) zed(8) zpool(5) systemd(1) systemd.target(5) systemd.special(7) systemd.mount(7)

August 24, 2020 OpenZFS