How do schedules and retention work, and how do I free storage space?

Add a schedule

1. Create the job. The Add schedule dialog opens by itself. For an existing job, add a schedule from its row.
2. Choose a Run trigger: Every hour, Daily, Weekly or Monthly.
3. Choose a Backup type: Incremental (the default) looks only at files the file watcher flagged as changed; Full checks every file of the account. See Incremental and deduplicated backups.
4. Set Day of week (weekly) or Day of month (monthly: 1 to 28, or Last day, the real last day of each month).
5. Set Hour and Minute, in the server's local time. Daily at 02:00 is the default.
6. Set Retention — points to keep: how many backups to keep for this schedule. The console suggests 14.
7. Choose Add schedule.

A job can have several schedules.

Schedule behavior to know

• Times are the server's local time. Around a daylight-saving change, a daily or monthly run can be an hour early or late.
• A new schedule first runs at its next occurrence. A missed run does not stack up; an overdue schedule runs once.
• An hourly schedule runs one hour after the last run started. It is not tied to the top of the hour, and the minute can drift.
• After a reboot, scheduled backups wait 30 minutes. Manual runs do not wait.
• The console offers Every hour, Daily, Weekly and Monthly. Every N hours, chosen weekdays and several days of a month can be set only through the API. There are no cron expressions.
• Only one backup runs per server at a time; a second job waits and is tried again every 2 minutes.

Retention

Retention keeps the newest N points per schedule. There is no daily, weekly and monthly rotation scheme: it is a count only.

• Retention runs every hour and only marks points deleted.
• A point is kept longer, up to 90 days, while it holds an account's newest backup.
• Manual runs keep at least 10 points, or as many as the job's most generous schedule keeps, and expire after 30 days.
• The console suggests 14 points to keep; the API default is 7.
• Safety copies taken before restores are kept 7 days, set in Settings → Server → Safety copies.

Why deleting backups does not free space right away

Removing a point only marks its data as unused. A cleanup that frees the space runs about once a day. To run it now, choose Reclaim space on the Backup points page.

• On a local destination, space comes back when the cleanup removes the objects.
• On remote destinations, a pack file is rewritten only when it is at least 6 hours old and less than 70% in use, so a mostly-used pack keeps its unused bytes for a long time. The console does not show how much is pending.
• If a point cannot be read, the cleanup skips that whole destination and frees nothing (a log line only).
• The cleanup never deletes the objects of a backup point it cannot find in the database. It keeps them and alerts "N backups on destination X are not in this server's database". Rescan destination imports them as points.
• You can also delete a single point manually from its detail page, then reclaim space.