Geo - Upgrading to GitLab 13.12 cause secondary nodes to redownload each LFS file
Summary
The problem is that the new state
column in the lfs_objects_registry
table introduced in 34aab0b7 is set to 0 by default which means pending
that causes Geo to redownload each LFS file.
Possible workaround
The current workaround is to jump into a Rails console on each secondary node and update the state
column properly with any of the following snippets:
- Migrate legacy sync state to the new
state
column (should take under a minute):
Geo::LfsObjectRegistry.where(state: 0, success: true).update_all(state: 2)
Edited by Michael Kozono