About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
The Bareos Director writes this line when an Incremental job looks in the catalogue for a Full backup of the same job, client and FileSet and finds none that counts. It then upgrades the job to a Full and carries on. The job can finish green, so the only symptom may be a night that took many times longer than usual.
The documented reasons are six: a first run, an edited FileSet, no successful Full on record (for example because every Full failed), a Full whose record was pruned, a job or client that was renamed, or a Full older than the job's maximum interval. Each one leaves a different trace in the catalogue, and you can read it before changing anything.
What the message means
The Bareos documentation says that for an Incremental, "If the Director cannot find a previous valid Full backup then the job will be upgraded into a Full backup." It lists what the Director looks for in the catalogue: the same Job name, the same Client name, the same FileSet, a previous Full, Differential or Incremental backup that "terminated normally", and one that "started no longer ago than Max Full Interval". If any of these does not hold, "the Director will upgrade the Incremental to a Full save."
The wording below comes from the Bareos source on the bareos-25 branch. Check it against your own log. The catalogue lookup (FindJobStartTime) first asks for the latest Full with the same job name, client and FileSet whose status is T or W. The Bareos catalogue documentation describes T as completed successfully and W as terminated with warnings. A Full that ended in E (errors) or f (fatal error) does not match.
For an Incremental that finds no such Full, the lookup sets the message "No prior Full backup Job record found." and the Director logs it as an information line. Then GetLevelSinceTime logs a second line that begins "No prior or suitable Full backup found in catalog." and ends with the words Doing FULL backup, and the job report shows the level as Full followed by the words upgraded from Incremental in brackets.
Two details help you read the log:
- The second line also appears when a Full exists but is older than Max Full Interval. In that case the first line is absent.
- A Differential with no Full does not print the first line. The same lookup reports "No Job record found" followed by the query text, and the job is upgraded the same way.
Likely causes
| Cause | What you see | How to confirm |
|---|---|---|
| First run of a new job, or a new client | First job ever for that name is a Full | list job=<job-name> shows no earlier runs |
| FileSet edited | Full follows the edit, once | The documentation: any change to Include or Exclude "constitutes a different FileSet"; the last Full belongs to the old one |
| No successful Full on record: every Full for this job, client and FileSet failed or was cancelled | The first line appears. A failed Full does not trigger it while an older Full with status T or W still exists |
No Full of this job has status T or W |
| Full record pruned | Full existed months ago, now gone from the list | Compare the Full's age with the retention periods in the Client and Pool resources |
| Job, client or FileSet renamed | Incrementals start over under the new name | Old name has history; new name has none |
| Full older than Max Full Interval | Second line only, no "No prior Full" line | Directive value compared with the date of the last Full |
Prove the cause before changing anything
These checks only read the catalogue. Run them in bconsole.
list jobs
list job=<job-name>
list joblog jobid=<jobid-of-the-upgraded-job>
list jobs, list job = <job-name> and list joblog jobid = <id> appear in the Bareos console documentation. Read the listing for the level and final status of each run, and find the last Full of this job. In the upgraded job's log, the two information lines above should appear.
Read the results this way:
- A Full with status
TorWexists, is recent, and the job name, client and FileSet are unchanged. The cause is not a missing Full. Check Max Full Interval next. - The last Full ended in an error or was cancelled. A failed Full never counts, but if an older Full with status
TorWexists, the lookup finds that one and no upgrade follows from this cause. Only when no Full of this job has statusTorWdoes the first line appear. - There is no Full under this job name at all. Look for a rename, a new job definition or a catalogue that was restored from an older copy.
- The Full is there but under a different FileSet. Someone edited the Include or Exclude lists after it ran.
To see the directives, run show all, which the console documentation describes as listing all resources, or read the Director configuration files. You need Max Full Interval (Job), Ignore File Set Changes (FileSet), Rerun Failed Levels (Job), and the Job Retention, File Retention and Auto Prune directives in the Client resource. The Director page marks those three Client directives "Since Version: deprecated", and it marks the Pool File Retention and Job Retention directives the same way. The page does not name a replacement for them; Volume Retention in the Pool is not marked deprecated.
Fix
Each of these is documented behaviour. The fix is to stop the surprise, not to suppress the message.
First run, or renamed job. Nothing is wrong. Schedule the Full out of business hours by setting the level to Full in the Schedule resource, which the documentation says overrides the job's own Level. For a rename, accept one Full under the new name.
Edited FileSet. The documentation says a change to the Include or Exclude lists forces the next backup to a Full "so that Bareos can guarantee that any additions or deletions are properly saved." Make FileSet edits shortly before a scheduled Full, and batch several edits into one.
Disruptive step: Ignore File Set Changes = yes stops the forced Full. The documentation says "We strongly recommend against setting this directive to yes, since doing so may cause you to have an incomplete set of backups." To back out, set it to no and run a Full.
Failed Full. Fix why the Full failed first. Then consider Rerun Failed Levels = yes (default no). The documentation says that when Bareos detects that a previous Full or Differential failed, "the current job level will be upgraded to the higher level." That saves a night of Incrementals that cannot be restored from. It also means any FileSet change triggers a rerun.
Risk: Rerun Failed Levels = yes adds extra Fulls or Differentials, each of which reads the whole data set or everything changed since the last Full. To back out, set it to no or remove the directive; Fulls already run cannot be undone, but they only add restore points.
Pruned Full. Pruning is meant to keep a current backup. The Bareos source comment for job pruning says it deletes a Job record only if there is a more current backup of the FileSet, and "we never delete the only Job record that contains a current backup of a FileSet." That protection depends on the job, FileSet and client still existing in the configuration. The same source treats records whose job, FileSet or client no longer exist in the configuration as deletable. Removing or renaming a resource can therefore allow its old Full to be pruned.
The Job Retention documentation also notes that the Job retention period "can actually be less than the value you specify here" when the Volume retention directive is set to a smaller duration, because the periods are applied independently and the smaller wins.
Max Full Interval. The directive says Incremental and Differential jobs are upgraded to Full once the most recent successful Full is older than the interval. If your schedule takes a Full weekly, set the interval longer than that gap, or a delayed Full will trigger an extra one.
Risk: changing Max Full Interval changes when Fulls are forced. A longer value means Fulls arrive less often and Incrementals depend on an older Full; a shorter value adds extra Fulls and reads. To back out, restore the previous value or remove the directive.
Confirm the fix
After the next scheduled run, list job=<job-name> should show an Incremental with status T, and the job log should not say upgraded from Incremental. For the FileSet case, confirm that one Full ran after the edit and that Incrementals follow it. Time the Incremental: it should be close to the size of one day's changes, not the whole data set.
Example (illustrative): job fs01-home takes a Full on Sunday and Incrementals Monday to Saturday. On Wednesday afternoon an administrator adds /srv/projects to the FileSet on fs01. The data in scope is now 1.2 TiB (1,319,413,953,331 bytes, or 1,228.8 GiB), and a normal Incremental is 18 GiB. Thursday's job is upgraded to a Full. It moves about 68 times the usual data (1,228.8 divided by 18 is 68.3). At a sustained 100 MB/s, 18 GiB takes about 3.2 minutes and 1.2 TiB takes about 3.7 hours (1,319,413,953,331 divided by 100,000,000 is 13,194 seconds). If the backup window is four hours, the job still fits but has almost no headroom.
What to watch afterwards
- Count upgraded Fulls per month. More than one a month on a weekly-Full schedule points to a FileSet being edited, failed Fulls or a short Max Full Interval.
- Watch the duration of Incrementals. A run that suddenly takes hours is an upgrade, even when the report is green.
- After any rename, check the first Full lands in a Pool with enough space.
- Keep the Full's record in the catalogue for as long as the Incrementals depend on it. The Bareos restore documentation warns that if File records of needed jobs are pruned, "the restore will be incomplete", and that Bareos does not detect this itself. Look for jobs with JobFiles at zero in the list of jobs selected for a restore.
Read next
Related articles you may want next.
- Backup & RecoveryReading a backup report: what success, warning, partial, failed and missed each mean for recoveryWhat backup report statuses mean for recovery: success, warning, partial, failed and missed, using AWS, Azure and Veeam documentation, with a triage rule.Read article
- Backup & RecoverySize Backup Retention Using Measured Storage GrowthBuild a capacity forecast from stored growth, copy formats and retention dependencies, with explicit headroom, recovery needs and uncertainty.Read article
- Backup & RecoveryGrandfather-father-son retention: counting the restore points a policy really keepsHow to count the restore points a 30 daily, 12 monthly backup policy keeps, where the gaps fall, and what to ask a provider to confirm in writing.Read article
Risks and limits
These commands have not been run or lab-tested. Message wording comes from the bareos-25 source branch and may differ in your release; other releases and the Bacula fork can word it differently. The Bacula 13.0.x and 15.0.x manuals describe the same level rules for Incremental jobs, but the exact log text comes from the Bareos source only.
Do not run a Full during business hours just to clear the message. A Full reads the whole data set, and on a busy file server it competes with users. Do not change retention or prune directives to make an upgrade go away: shortening retention can delete the Full that the next restore needs.
Sources
- Director Configuration (Job Level, Max Full Interval, Ignore File Set Changes, Rerun Failed Levels, Client Job Retention) — Bareos project, footer: Revision November 06, 2026, accessed 17 November 2026
- Catalog Tables (JobStatus codes) — Bareos project, footer: Revision November 06, 2026, accessed 17 November 2026
- The Restore Command — Bareos project, footer: Revision November 06, 2026, accessed 17 November 2026
- Bareos Console (list commands) — Bareos project, footer: Revision November 06, 2026, accessed 17 November 2026
- Bareos source, core/src/cats/sql_find.cc (branch bareos-25) — Bareos project (GitHub), date not shown, accessed 17 November 2026
- Bareos source, core/src/dird/job.cc (branch bareos-25) — Bareos project (GitHub), date not shown, accessed 17 November 2026
- Bareos source, core/src/dird/ua_prune.cc (branch bareos-25) — Bareos project (GitHub), date not shown, accessed 17 November 2026
- Bareos source, core/src/dird/backup.cc (branch bareos-25; the job report line "Backup Level") — Bareos project (GitHub), date not shown, accessed 17 November 2026
- Director Configuration (Bacula 15.0.x) — Bacula project documentation, date not shown, accessed 17 November 2026
- Director Configuration (Bacula 13.0.x) — Bacula project documentation, date not shown, accessed 17 November 2026
This article was originally published on .