Glossary
What you will learn
- What the words in this book mean.
Terms are in alphabetical order.
ACL: access control list: the rules saying who may read, write or execute a file. FreeBSD ZFS uses NFSv4 ACLs.
alert: a message sent when a host turns bad on two checks in a row; see chapter 16.
ashift: the sector size of a vdev as a power of two (12 is 4096 bytes), fixed when the vdev is created; chapter 2.
ARC: adaptive replacement cache: ZFS's memory cache of recently used data; shown in the host panel.
backup: a separate copy of your data, elsewhere, that you can restore from by itself; chapter 15.
bectl: FreeBSD's tool for boot environments.
browse: looking inside a snapshot's folders (f in Snapshots) with each entry's type, size, dates, mode and owner; names and metadata only, never contents; other hosts need the off-by-default browse sudoers feature; chapter 10.
boot environment: a bootable copy of the operating system kept as a ZFS clone, so a bad upgrade can be undone; chapter 13.
restore as a copy: getting one file back out of a snapshot by copying it to a new name next to the original (R in the snapshot browser); it never overwrites or replaces anything; needs the off-by-default restore sudoers feature on other hosts; chapter 10.
catch up: running a missed scheduled job once when the daemon starts.
checksum: a fingerprint ZFS stores with every block to detect silent corruption.
clone: a writable copy of a snapshot that shares its blocks.
compression: ZFS can compress data as it writes it; compression=on means lz4, the default for new pools on OpenZFS 2.2 and later, and zstd compresses harder for more CPU.
copy-on-write: ZFS never overwrites a block in place; it writes a new one and then moves the pointer, which is what makes snapshots free.
cron: the five-field schedule notation used by jobs.
daemon: a program that runs in the background; zfsctl daemon runs the jobs and the health monitor.
dataset: a filesystem or volume inside a pool; chapter 9.
dedup: deduplication: storing identical blocks once; the table that finds them grows with the number of unique blocks, so it is costly in memory and rarely worth it; chapter 9.
delegation: letting an ordinary user perform certain ZFS actions, with zfs allow.
destroy: to delete a pool, dataset, snapshot or boot environment for good.
disk: a physical storage device such as ada1; chapter 11.
Disks: the section that lists the host's disks and builds pools from them.
ENV: an environment variable; the notify hook receives its details this way.
escalation: an alert that gets worse, from yellow to red.
export: to detach a pool so it can be moved to another machine (zpool export).
feature flag: a pool capability. Once enabled it cannot be disabled, and one that is in use (active) can stop older software from importing the pool; chapter 8. fragmentation: how scattered a pool's free space is; high values slow writes.
FreeBSD: the operating system zfsctl manages; chapter 3.
guest access: an SMB share anyone can reach without a password.
health: a pool's state: ONLINE, DEGRADED, FAULTED and so on.
heartbeat: a file the daemon rewrites every second so the TUI can tell it is running.
hold: a lock ZFS puts on a snapshot so it cannot be destroyed until the hold is released; zfsctl's own tag is zfsctl; chapter 10.
hook: the notify_command that runs for alerts and failed jobs; the way to reach Telegram, ntfy and other services.
host: a machine zfsctl manages; the hub itself is local; chapter 7.
hub: the one machine that runs zfsctl and the daemon, and reaches the others over ssh.
import: to attach a pool to the running system (zpool import).
incremental: a send that carries only what changed since a common snapshot.
jail: FreeBSD's lightweight container.
job: a scheduled task of the daemon: scrub, snapshot, replicate or backup; chapter 14.
keep: how many of a job's snapshots to retain; for backups, restic's retention words.
L2ARC: an optional second-level read cache on a fast disk.
lz4: the usual ZFS compression algorithm: fast, and often a net speed-up.
mirror: a vdev of two or more disks holding the same data.
mountpoint: where a dataset appears in the directory tree.
NFS: the Unix network filesystem; chapter 12.
offline: a disk taken out of service on purpose.
pool: one or more vdevs presented as a single store of space; chapter 8.
prefix: the start of the name of snapshots a job makes and uses, such as hourly.
property: a setting of a pool or dataset, such as compression or quota.
quota: a limit on the space a dataset may use.
raidz: a vdev with parity: raidz1 survives one failed disk, raidz2 two, raidz3 three. (Like RAID 5/6)
rclone: a tool that talks to dozens of cloud storage services; restic can use it as a repository.
recordsize: the block size ZFS uses for files in a dataset.
redundancy: extra copies or parity that let a pool survive a failed disk.
refreservation: space guaranteed to a dataset itself, not counting its snapshots.
replication: copying a dataset and its snapshots to another place with zfs send and receive; chapter 14.
repository: where restic keeps a backup: a path, an sftp address or an rclone remote.
reservation: space guaranteed to a dataset.
resilver: the repair after a disk is replaced or returns: data is rebuilt onto it.
restic: the open-source backup tool zfsctl's backup jobs run.
rollback: returning a dataset to a snapshot, discarding everything after it.
root: the administrator account; zfsctl runs as root on the hub.
rtt: round-trip time of a host's last health probe.
Samba: the software that provides SMB sharing on FreeBSD.
scrub: reading all data and checking it against the checksums, repairing from redundancy; chapter 8.
send and receive: zfs send writes a snapshot as a stream; zfs receive rebuilds it elsewhere.
share: a dataset made available to other machines over NFS or SMB; chapter 12.
sharenfs: the dataset property that holds NFS export options.
SMB: the Windows file-sharing protocol, served by Samba.
snapshot: a read-only picture of a dataset at one moment; chapter 10.
source: where a property's value comes from: local, inherited or default.
stripe: a pool with no redundancy: if any disk fails, everything is lost.
sudoers: the file that says which commands an account may run as root; chapter 5.
tag: your own label for a host.
temporary boot: using a boot environment for the next boot only.
TRIM: telling an SSD which blocks are free; the pool property autotrim does it automatically; chapter 8.
theme: the colour scheme of the screen.
vdev: a virtual device: a disk, a mirror or a raidz group, from which pools are built; chapter 8.
volume: a ZFS block device, used for virtual machine disks or iSCSI.
zfs command: the command-line tool for datasets and snapshots.
ZFS: the filesystem and volume manager that zfsctl manages; chapter 2.
zpool command: the command-line tool for pools.