GC doesn't work in the next run after click on stop button

[PrzemekZglinicki]

I want to run Garbage Collection form web UI. I started it but in the same time I read somewhere in the internet that it removed all images for someone, so I decided to stop it, pull locally the most important images to have backups and run it again. In the next run GC didn't detect any blobs and manifests to remove.

GC didn't removed any blobs and manifests in the first run which was stopped, and didn't detect any blobs and manifests to remove in the next run.

Expected behavior and actual behavior:
Stop button shouldn't impact next runs of the GC.

Steps to reproduce the problem:
Set up retention policy, observe that size showed in the general projects and inside project are different - due to run of retention policy and remaining garbage blobs and manifests, run GC, stop it in the meantime, run GC again.

Versions:
Please specify the versions of following systems.

• harbor version: 2.12.1
• docker engine version: 27.1.2
• docker-compose version: 1.29.21

Additional context:
harbor_log_and_config.zip

[stonezdj]

From the jobservice logs, we can find two GC execution did free space.

Jan 14 10:34:10 misc jobservice[938]: 2025-01-14T10:34:10Z [INFO] [/jobservice/job/impl/gc/garbage_collection.go:480]: 14 blobs and 3 manifests are actually deleted
Jan 14 10:34:10 misc jobservice[938]: 2025-01-14T10:34:10Z [INFO] [/jobservice/job/impl/gc/garbage_collection.go:481]: The GC job actual frees up 140 MB space.

Jan 14 11:09:12 misc jobservice[938]: 2025-01-14T11:09:12Z [INFO] [/jobservice/job/impl/gc/garbage_collection.go:480]: 38 blobs and 10 manifests are actually deleted
Jan 14 11:09:12 misc jobservice[938]: 2025-01-14T11:09:12Z [INFO] [/jobservice/job/impl/gc/garbage_collection.go:481]: The GC job actual frees up 2515 MB space. 

[wy65701436]

If the issue you're referring to is this one, it has already been resolved. This issue only occurred in specific edge cases, as explained in this comment.

Regarding the second execution, based on the logs you attached, the first execution that was stopped has already completed successfully and removed all the candidates. That’s why there are no remaining candidates to be removed. Please try creating more candidates and then retry the GC process.