How do I run a backup and read the result?

Run a backup now

In Backup jobs, choose the Run now icon on the job. A manual run is always a Full backup: every file is checked and only changed or new files are read.
To back up only one account, open Accounts and choose Backup now on its row.
The job's status changes to Running and shows the elapsed time. There is no progress bar for backups.
When it finishes, open Backup points. Your run is listed with its status, size and new data. Choose it for details.
Choose Verify now in the point's detail to check that every stored object is still on the destination. See Verify your backups.

To follow a run in the logs:

tail -f /var/log/backupwall/worker.log

Status words

Jobs show Running, Success, Failed, Pending, Disabled and Warnings. A point that backed up only some accounts, or finished with problems, reads Partial: N items need attention in the jobs list (hover for the reason), with a Partial filter on the list and "Needs attention" on the dashboard.

What makes a point Partial or carry a warning

A database dump fails after its retry, or its upload fails: the point is Partial, naming the database and why. A partial point can still be restored; the failed database is marked missing in it. A dump that ends early with exit code 0 (a cut gzip stream, or no -- Dump completed last line) is caught and retried.
An account's database list could not be read from the panel or the database server: named on the point, and the point is Partial.
A file exists but could not be read or stored: the point shows a Warnings badge with "Files not backed up (they exist but could not be read or stored; ...)" and up to five examples per account. The new point keeps the older stored version of that file, if there is one, and the next run tries again.
A section could not be captured (DNS zones, SSL certificates, FTP logins, cron jobs, mail settings): the point records which sections it captured and names any read that failed, so a failed read never looks like "the account has none".
PostgreSQL, symlink targets and files outside the home are named as not backed up where they apply.

A file deleted while the backup reads it is not reported. The message is cut at 500 characters, so on a busy point it may be shortened.

How warnings reach you

Email shows "completed with errors". On Slack, Discord and webhook channels the alert comes through the Backup failed event, titled "Backup job 'X' partial". Subscribe to that event if you want warnings. See Alerts and notifications.

Interrupted and long runs

If the worker restarts or crashes mid-run (an upgrade, a reboot), the point is marked failed with "Interrupted: the worker process restarted or crashed before this run finished". Nothing restarts it by itself; the job's next scheduled run, or Run now, takes over and reuses what was already uploaded, so it sends only what is missing.
There is no fixed time limit. A run is kept alive while it reads or writes data. A run that does nothing for 2 hours is stopped (it gets 10 more minutes to stop, then it is killed); a backup is then tried again, up to three times. A run is stopped after 7 days at the latest.
While a long backup runs, the server's other jobs wait for the slot.

Where to look

Backup points lists every run with its status and message.
Audit log shows who did what and when.
/var/log/backupwall/worker.log shows why a backup failed.