A collection of scripts for working with ZFS Snapshots. Most assume ZFS as hosted by FreeNAS (snapshot naming conventions, API support), but should be useable on any ZFS system.
Similar in behavior to Apple's TimeMachine, the default behavior is to keep hourly snapshots for the last day, daily for the last week, and weekly thereafter. Extending beyond TimeMachine, the argument syntax allows customizing the number of snapshots kept at each interval, as well as defining additional buckets of arbitrary interval lengths.
The goal here is to remove any snapshots that are of zero size, meaning the snapshot holds no unique changes. If the blocks that have changed in a snapshot are referenced by another snapshot, it will report zero size. If a dataset contains multiple snapshots with no unique blocks, the snapshots can be pruned. This will not free any blocks, but will declutter the list of snapshots.
NOTE: It is possible for snapshots to report zero unique blocks if the changed blocks are referenced by multiple snapshots. It is therefore important to only remove one snapshot at a time from a given dataset and then scan for additional empty snapshots.
By default, destroy all snapshots with an 'auto' prefix, except for the most recent snapshot. Options are available for changing the prefix, dryrun, and verbose output.
NOTE: The script currently never actually destroys any snapshots. If the destroy commands look acceptable, you can pipe them to another shell to perform the actions. This works with dryrun and verbose modes as well.
snap-strip.py tank tank/dataset | bash
See also README-tmsnap
This is a script to be run on a macOS machine. It will scan the system.log and act on messages from "backupd" to create a new snapshot on a FreeNAS machine that is providing the TimeMachine target.
It is not uncommon for TimeMachine to corrupt the TimeMachine volume sparsebundle, especially when connected over a wireless network. Corruption can also result from shutting down or sleeping the Mac. In most cases these will present when TimeMachine checks the integrity of the volume and reports that it needs to start over.
These issues can sometimes be resolved by stopping any current backups, disabling the TimeMachine service to prevent another backup from starting, and rolling back the TimeMachine dataset to an earlier snapshot before the issues started. Rolling back like this isn't guaranteed to resolve issues, and may lead to missing backups if the macOS machine does not detect that it needs perform a full scan prior to backing up. In the worst case, the dataset can be rolled back to an empty state if such a snapshot exists.
NOTE: When rolling back the TimeMachine dataset to an empty state, it's likely that TimeMachine will report that the volume identity has changed.
WARNING: It is also possible for FreeNAS errors to interfer with the AFP service in a way that mimics sparsebundle errors. Before rolling back any datasets, try restarting the AFP service on FreeNAS. Also look for any processes that are in states of "Uninteruptible Sleep" and "pages locked into memory" (D,L respectively). This may indicate a deadlock that will not be recovered by restarting the AFP service and will require rebooting the FreeNAS machine.