View Issue Details

IDProjectCategoryView StatusLast Update
0001592bareos-coredirectorpublic2024-03-28 15:26
Reporterhostedpower Assigned Tobruno-at-bareos  
PrioritynormalSeveritymajorReproducibilityalways
Status closedResolutionfixed 
PlatformLinuxOSDebianOS Version12
Product Version22.1.3 
Summary0001592: Rerun failed consolidation group purges all jobs except latest
DescriptionHi,


We sometimes see failed Consolidate jobs (for the always incremental scheme). These jobs work fine normally, and all is consolidated properly in normal circumstances.

However when we have a failed job for some host for whatever reason, and we re-run it, ALL restore points are lost (except the very latest one).

This is very dangerous, since when it's done, we lose all but the latest restore point.

I believe this can also be tested / reproduced by just re-running a consolidation job which succeeded.

This bug is there long time (version 18 or longer back) but was never reported before. Since it's so dangerous and not fixed, we finally decided to report it here :)
TagsNo tags attached.

Activities

bruno-at-bareos

bruno-at-bareos

2024-01-23 17:04

manager   ~0005715

rerun should never never never (repeat after me: NEVER) be used with a failed job prepared by consolidation, as it doesn't do what consolidation is doing (create a precise list of jobid to consolidate) rerun don't do that, so you get a virtual full of everything.

What you have to if you want to rerun a consolidation, is rerun a consolidation job which will create the "consolidated job" correctly, which is documented.

We have a task to make the rerun without a jobid list in case of virtual full automatically canceled, but I can't promise when this will be developped.
hostedpower

hostedpower

2024-01-23 17:06

reporter   ~0005716

Haha that is clear now, but the error is easily made by anyone if it's not protected. Thx for confirming this is a known "bug" at least :)
bruno-at-bareos

bruno-at-bareos

2024-03-18 16:23

manager   ~0005837

Fix in progress at https://github.com/bareos/bareos/pull/1738
hostedpower

hostedpower

2024-03-18 16:40

reporter   ~0005838

Cool, thx a lot, will be a lot safer :)
bruno-at-bareos

bruno-at-bareos

2024-03-28 15:26

manager   ~0005889

While https://github.com/bareos/bareos/pull/1738 is not yet merge and backported we will take care of this.
I'm closing this bug before the changes of Mantis being ReadOnly and the opening of github issues.

Issue History

Date Modified Username Field Change
2024-01-23 15:32 hostedpower New Issue
2024-01-23 17:04 bruno-at-bareos Note Added: 0005715
2024-01-23 17:06 hostedpower Note Added: 0005716
2024-01-24 11:23 bruno-at-bareos Assigned To => bruno-at-bareos
2024-01-24 11:23 bruno-at-bareos Status new => acknowledged
2024-03-18 16:23 bruno-at-bareos Note Added: 0005837
2024-03-18 16:40 hostedpower Note Added: 0005838
2024-03-28 15:26 bruno-at-bareos Status acknowledged => closed
2024-03-28 15:26 bruno-at-bareos Resolution open => fixed
2024-03-28 15:26 bruno-at-bareos Note Added: 0005889