System Backup Policies
System backup policies (Storage > System Backup Policies) take backups automatically on a schedule, without users or per-server schedules being involved. A policy covers a set of servers, runs on a cron schedule, and cleans up after itself with retention rules.
A policy backs up one of two things, chosen by its Backup Kind: the files of every covered server, or the contents of every database instance of every covered server. The kind decides what the policy can be scoped to, so it is fixed when the policy is created.
The list shows each policy's ID, Name, Schedule (the cron expression), Backups (how many it has taken), Last Run, and Created. Next to the name sit its badges:
- Server or Database for the backup kind.
- Retention (hover it to see the rules, the same tooltip backup groups use) or No retention rules.
- Disabled on disabled policies.
Since Panel 1.2.9, Last Run shows the state of the policy's most recent run and when it started. It replaces the Run pending badge older versions showed next to the name.
Use the search box to filter.

Creating a Policy
Click Create in the top right.

| Field | Description |
|---|---|
| Name / Description | Display name and optional free text. |
| Backup Kind | Server archives the files of each covered server. Database dumps each covered database instance. The choice is locked once the policy exists, because it decides which scopes apply. |
| Backup Configuration | Which backup configuration the policy's backups are written to. Defaults to Inherit from Server, the same resolution a manual backup would use. |
| Schedule | "Cron expression (with seconds) that determines when backups are taken, in UTC." The field has a helper popover for building the expression segment by segment, describes the current expression in plain English underneath, and previews when the next run would be. |
| Retention | Six rules deciding which of the policy's backups survive: Keep latest, Keep all within days, Keep daily, Keep weekly, Keep monthly, and Keep yearly, each off when empty or 0. They work as they do for backup groups, and the field has a helper popover describing them. The rules apply per server, and for a database policy separately to each database instance. |
| Parallelism | "Maximum number of backups this policy runs at the same time on a single node." Database policies apply the same number per database agent host instead, since the dump is read from the host rather than the node. |
| Enabled | "Disabled policies keep their backups but do not take new ones." |
An existing policy also offers Run Now (after a Confirm Manual Run dialog) and Delete. Run Now starts a manual run covering every target that has no backup from this policy in progress. The button shows a spinner while the latest run still has targets waiting, and the panel refuses Run Now while any open run of the policy still has targets waiting ("a run of this system backup policy is still in progress"). It also works on a disabled policy.

What Is Covered
A server policy has Locations, Nodes, and Servers tabs; a database policy has Locations, Database Agent Hosts, and Servers. Each tab has an Add button and right-click Remove.
For a server policy, a server is covered when it matches any of the three: it sits in an assigned location, runs on an assigned node, or is assigned directly.
For a database policy, a database instance is covered when its database agent host is assigned directly, when that host belongs to an assigned location, or when the instance's server is assigned directly.
Covered is not the same as backed up on every run. A scheduled run only picks up targets that were covered (and, for servers and instances, existed) before the cron time it belongs to, so a server added to the policy afterwards waits for the next scheduled run or a Run Now. A target whose previous backup from this policy is still in progress is left out of the new run.
Inside a run, some conditions only hold a target back. The panel retries it every two minutes and records the reason on the target. Such a target keeps the run open until it clears, the policy's next scheduled run starts, or the run is cancelled:
- The server has any status, including install failed or restore failed, or is being transferred, or the database instance is restoring a backup.
- The node is in maintenance mode, or for a database policy, the database agent host is.
- The backup configuration is in maintenance mode.
- For a database policy, the database instance isn't running or its database agent can't be reached, since a dump is taken through the running database.
- The panel could not collect the server's backup metadata.
Other conditions skip the target for that run straight away: the server or instance was removed, no backup configuration applies to it, or (for a database policy) the backup configuration resolves to btrfs or zfs, which cannot store dumps.



Runs
Since Panel 1.2.9, every time the schedule fires the panel starts a run for the policy, and Run Now starts a manual one. The Runs tab lists them newest first, with Started, Trigger (Scheduled or Manual), Status, Progress, and Completed. The table refreshes every few seconds while a run is still open.

| Status | Meaning |
|---|---|
| Running | The run still has targets waiting or backups in progress. |
| Cancelling | The run was cancelled, and backups it already started are still finishing. |
| Cancelled | The run was cancelled and has finished. |
| Succeeded | Every target got a successful backup. |
| Partially Succeeded | Some targets got a successful backup, others failed or were skipped. |
| Failed | No target got a successful backup. |
| Nothing to Back Up | The run had no targets. |
Progress is a bar split into successful, failed, skipped, and running targets (hover a segment for its count), with the number of finished targets out of the total next to it.
Right-click a run, or use its menu, for View Targets and Cancel Run. View Targets lists the run's servers (and, for a database policy, a Source column with the database instance) with Node, Status, Reason, and Updated, searchable by server or instance name and filterable by stored status: Pending, Dispatched, Skipped, or Failed. A target reads Pending while it waits, Running while its backup is in progress, then Successful, Failed, or Skipped. A backup that has not finished within a day counts as failed. The Failed filter only finds targets whose backup could not be created; a backup that started and then failed is listed under Dispatched.

The reasons a target shows are:
| Reason | When |
|---|---|
| Server busy (installing, restoring or transferring) | The server, or the database instance, had a status that blocks backups. |
| Node in maintenance | The server's node is in maintenance mode. |
| Database agent host in maintenance | The instance's host is in maintenance mode. |
| Backup configuration in maintenance | The backup configuration is in maintenance mode. |
| Database instance not running | The instance's container was not running, or its database agent could not be reached. |
| Could not collect backup metadata | The panel could not gather the server's backup metadata. |
| No backup configuration | No backup configuration applies to the server. |
| Backup configuration does not support database backups | A database policy resolved to a btrfs or zfs configuration. |
| Removed before it was backed up | The server or instance was deleted while the run was open. |
| Backup could not be created | Creating the backup failed. The target counts as failed. |
| Policy disabled | The policy was disabled while a scheduled run still had targets waiting. |
| Next scheduled run started first | The next run started before this target got its turn. |
| Run cancelled | The run was cancelled before this target got its turn. |
When the next scheduled run supersedes a target that was held back, the target keeps the reason it was last held back, so a target that waited out a whole run on a node in maintenance reads "Node in maintenance" rather than "Next scheduled run started first". Cancelling the run sets "Run cancelled" and disabling the policy sets "Policy disabled" in its place.
Cancel Run (requires system-backup-policies.update) asks for confirmation, then skips every target still waiting. Backups that already started keep running until they finish. Disabling a policy also skips the waiting targets of its scheduled runs, with "Policy disabled", but doesn't mark the run cancelled, so it ends as Partially Succeeded or Failed. A manual run keeps going.
Within a run, targets are handled oldest previous attempt first, and at most Parallelism backups run at once per node (per database agent host for a database policy). If the panel was offline over several scheduled times, only the most recent one gets a run when it comes back.
The panel keeps the last 200 finished runs of each policy and the target lists of the last 20, pruning older ones every hour.
Backups
The Backups tab lists every backup the policy has taken, searchable: Name, Server, Node, Checksum, Size, Files, Retention, and Created, each row tagged with a SYSTEM badge and carrying a download action. A database policy adds a Kind column, and its rows report no file count because a dump is a single file.
Policy backups do not count towards each server's backup limit and no backup group touches them; the policy's own Retention rules manage them instead, re-checked once an hour. Unlocked failed attempts are cleaned up 24 hours after they finish.
When there are failed backups, a Delete Failed button appears above the table (requires nodes.backups), showing how many it would remove. It asks for confirmation, keeps locked backups and any whose configuration is in maintenance, and runs in the background. A Force switch removes them even when the configuration is missing or the remote storage is unreachable, at the risk of leaving orphaned files behind.
A backup whose deletion still fails after the panel's eight automatic attempts is marked as a failed deletion. Since Panel 1.2.8, when there are any, a Retry Failed Deletions button with their count sits next to it (requires system-backup-policies.backups and nodes.backups). It skips backups whose configuration is in maintenance, runs in the background, and has the same Force switch.

Deleting a Policy
Deleting asks whether to also delete the policy's backups (Do you want to delete backups created by this policy?):
- Switch on: "All backups created by this policy will be permanently deleted from their storage backends."
- Switch off: "Backups created by this policy will become regular server backups. They will count towards each server backup limit and follow standard rotation." Released backups belong to no group, so no retention rule applies to them afterwards.
INFO
The buttons on these pages follow the system-backup-policies.* admin permission keys (create, read, update, delete, backups). The Runs tab and its targets need read, and Run Now and Cancel Run need update. See the Permissions Reference.