Generic Options

The following arguments have the same meaning over several actions:

-c <conf>

Use the specified file as the configuration file. The default file, if none is specified is /usr/local/etc/zetaback.conf. The prefix of this file may also be specified as an argument to the configure script.

-d

Enable debugging output.

-n

Don't actually perform any remote commands or expunging. This is useful with the -d argument to ascertain what would be done if the command was actually executed.

-t <timestamp>

Used during the restore process to specify a backup image from the desired point in time. If omitted, the command becomes interactive. This timestamp is a UNIX timestamp and is shown in the output of the -s and -sx actions.

-rhost <host>

Specify the remote host that is the target for a restore operation. If omitted the command becomes interactive.

-rzfs <zfs>

Specify the remote ZFS filesystem that is the target for a restore operation. If omitted the command becomes interactive.

-h <host>

Filters the operation to the host specified. If <host> is of the form /pattern/, it matches 'pattern' as a perl regular expression against available hosts. If omitted, no limit is enforced and all hosts are used for the action.

-z <zfs>

Filters the operation to the zfs filesystem specified. If <zfs> is of the form /pattern/, it matches 'pattern' as a perl regular expression against available zfs filesystems. If omitted, no filter is enforced and all zfs filesystems are used for the action.

Actions

-v

Show the version.

-l

Show a brief listing of available backups.

-s

Like -l, -s will show a list of backups but provides additional information about the backups including timestamp, type (full or incremental) and the size on disk.

-sx

Shows an extended summary. In addition to the output provided by the -s action, the -sx action will show detail for each availble backup. For full backups, the detail will include any more recent full backups, if they exist. For incremental backups, the detail will include any incremental backups that are more recent than the last full backup.

-sv

Display all backups in the current store that violate the configured backup policy. This is where the most recent full backup is older than full_interval seconds ago, or the most recent incremental backup is older than backup_interval seconds ago.

If, at the time of the most recent backup, a filesystem no longer exists on the server (because it was deleted), then backups of this filesystem are not included in the list of violators. To include these filesystems, use the -svv option instead.

-svv

The violators summary will exclude backups of filesystems that are no longer on the server in the list of violators. Use this option to include those filesystems.

--files

Display the on-disk file corresponding to each backup named in the output. This is useful with the -sv flag to name violating files. Often times, violators are filesystems that have been removed on the host machines and zetaback can no longer back them up. Be very careful if you choose to automate the removal of such backups as filesystems that would be backed up by the next regular zetaback run will often show up as violators.

-a

Performs an archive. This option will look at all eligible backup points (as restricted by -z and -h) and move those to the configured archive directory. The recommended use is to first issue -sx --files then carefully review available backup points and prune those that are unneeded. Then invoke with -a to move only the remaining "desired" backup points into the archives. Archived backups do not appear in any listings or in the list of policy violators generated by the -sv option. In effect, they are no longer "visible" to zetaback.

-b

Performs a backup. This option will investigate all eligible hosts, query the available filesystems from the remote agent and determine if any such filesystems require a new full or incremental backup to be taken. This option may be combined with the -x option (to clean up afterwards.)

-ff

Forces a full backup to be taken on each filesystem encountered. This is used in combination with -b. It is recommended to use this option only when targeting specific filesystems (via the -h and -z options.) Forcing a full backup across all machines will cause staggered backups to coalesce and could cause performance issues.

-fi

Forces an incremental backup to be taken on each filesystem encountered. This is used in combination with -b. It is recommended to use this option only when targeting specific filesystems (via the -h and -z options.) Forcing an incremental backup across all machines will cause staggered backups to coalesce and could cause performance issues.

-x

Perform an expunge. This option will determine which, if any, of the local backups may be deleted given the retention policy specified in the configuration.

-r

Perform a restore. This option will operate on the specified backup and restore it to the ZFS filesystem specified with -rzfs on the host specified with the -rhost option. The -h, -z and -t options may be used to filter the source backup list. If the filtered list contains more than one source backup image, the command will act interactively. If the -rhost and -rzfs command are not specified, the command will act interactively.

When running interactively, you can choose multiple filesystems from the list using ranges. For example 1-4,5,10-11. If you do this, zetaback will enter multi-restore mode. In this mode it will automatically select the most recent backup, and restore filesystems in bulk.

In multi-restore mode, you have the option to specify a base filesystem to restore to. This filesystem will be added as a prefix to the original filesystem name, so if you picked a prefix of data/restore, and one of the filesystems you are restoring is called data/set/myfilesystem, then the filesystem will be restored to data/restore/data/set/myfilesystem.

Note that, just like in regular restore mode, zetaback won't create intermediate filesystems for you when restoring, and these should either exist beforehand, or you should make sure you pick a set of filesystems that will restore the entire tree for you, for example, you should restore data as well as data/set before restoring data/set/foo.

Storage Classes

In addition to the default and host stanzas, the configuration file can also contain 'class' stanzas. Classes allow you to override settings on a per-filesystem basis rather than a per-host basis. A class stanza begins with the name of the class, and has a setting 'type = class'. For example:

  myclass {
    type = class
    store = /path/to/alternate/store
  }

To add a filesystem to a class, set a zfs user property on the relevant filesystem. This must be done on the server that runs the zetaback agent, and not the zetaback server itself.

  zfs set com.omniti.labs.zetaback:class=myclass pool/fs

Note that user properties (and therefore classes) are are only available on Solaris 10 8/07 and newer, and on Solaris Express build 48 and newer. Only the server running the agent needs to have user property support, not the zetaback server itself.

The following settings can be included in a class stanza. All other settings will be ignored, and their default (or per host) settings used instead:

• store
• full_interval
• backup_interval
• retention
• dataset_backup
• violator_grace_period

Settings

The following settings are valid in both the default and host scopes:

store

The base directory under which to keep backups. An interpolated variable '%h' can be used, which expands to the hostname. There is no default for this setting.

archive

The base directory under which archives are stored. The format is the same as the store setting. This is the destination to which files are relocated when issuing an archive action (-a).

agent

The location of the zetaback_agent binary on the host. There is no default for this setting.

time_format

All timestamps within zetaback are in UNIX timestamp format. This setting provides a string for formatting all timestamps on output. The sequences available are identical to those in strftime(3). If not specified, the default is '%Y-%m-%d %H:%M:%S'.

backup_interval

The frequency (in seconds) at which to perform incremental backups. An incremental backup will be performed if the current time is more than backup_interval since the last incremental backup. If there is no full backup for a particular filesystem, then a full backup is performed. There is no default for this setting.

full_interval

The frequency (in seconds) at which to perform full backups. A full backup will be performed if the current time is more than full_interval since the last full backup.

retention

The retention time (in seconds) for backups. This can be a simple number, in which case all backups older than this will be expunged.

The retention specification can also be more complex, and consist of pairs of values separated by a comma. The first value is a time period in seconds, and the second value is how many backups should be retained within that period. For example:

retention = 3600,4;86400,11

This will keep up to 4 backups for the first hour, and an additional 11 backups over 24 hours. The times do not stack. In other words, the 11 backups would be kept during the period from 1 hour old to 24 hours old, or one every 2 hours.

Any backups older than the largest time given are deleted. In the above example, all backups older than 24 hours are deleted.

If a second number is not specified, then all backups are kept within that period.

Note: Full backups are never deleted if they are depended upon by an incremental. In addition, the most recent backup is never deleted, regardless of how old it is.

This value defaults to (14 * 86400), or two weeks.

compressionlevel

Compress files using gzip at the specified compression level. 0 means no compression. Accepted values are 1-9. Defaults to 1 (fastest/minimal compression.)

ssh_config

Full path to an alternate ssh client config. This is useful for specifying a less secure but faster cipher for some hosts, or using a different private key. There is no default for this setting.

dataset_backup

By default zetaback backs zfs filesystems up to files. This option lets you specify that the backup go be stored as a zfs dataset on the backup host.

offline

Setting this option to 1 for a host will mark it as being 'offline'. Hosts that are marked offline will not be backed up, will not have any old backups expunged and will not be included in the list of policy violators. However, the host will still be shown when listing backups and archiving.

violator_grace_period

This setting controls the grace period used when deciding if a backup has violated its backup window. It is used to prevent false positives in the case where a filesystem is still being backed up. For example, if it is 25 hours since the last daily backup, but the daily backup is in progress, the grace period will mean that it is not shown in the violators list.

Like all intervals, this period is in seconds. The default is 21600 seconds (6 hours).

Global Settings

The following settings are only valid in the default scope:

process_limit

This setting limits the number of concurrent zetaback processes that can run at one time. Zetaback already has locks on hosts and datasets to prevent conflicting backups, and this allows you to have multiple zetaback instances running in the event a backup takes some time to complete, while still keeping a limit on the resources used. If this configuration entry is missing, then no limiting will occur.

Uniform hosts

This config results in backups stored in /var/spool/zfs_backups, with a subdirectory for each host. Incremental backups will be performed approximately once per day, assuming zetaback is run hourly. Full backups will be done once per week. Time format and retention are default.

  default {
    store = /var/spool/zfs_backups/%h
    agent = /usr/local/bin/zetaback_agent
    backup_interval = 83000
    full_interval = 604800
  }
  host1 {}
  host2 {}

Non-uniform hosts

Here, host1's and host2's agents are found in different places, and host2's backups should be stored in a different path.

  default {
    store = /var/spool/zfs_backups/%h
    agent = /usr/local/bin/zetaback_agent
    backup_interval = 83000
    full_interval = 604800
  }
  host1 {
    agent = /opt/local/bin/zetaback_agent
  }
  host2 {
    store = /var/spool/alt_backups/%h
    agent = /www/bin/zetaback_agent
  }