With the farm-vet feature implemented with Bitbucket DC 6.7 for mirrors, the Rest API to force sync repository changes from upstream/primary is no longer applicable. The farm-vet expected to self-heal by detecting corrupted objects and fix it by deleting and sync from upstream/primary. The far-vet is able to detect the corrupted object but unable to repair the corrupted object. Manual intervention is required by removing the project from the mirror sync list. Then restarting the mirror node and allow it to sync entire mirrored content from upstream/primary.
As a Bitbucket DC Sys. Admin, I would like to see, farm-vet to be able to force resync a repository/project without having to restart the entire mirror. Prior to 6.7 when we had an issue with a corrupt repository, we would remove the project and readd it or there was some API call. This is no longer possible 6.7+. And we have to restart the entire mirror which means downtime. Of course, having a mirror-farm setup is going to help minimize the downtime and is something we're looking to do soon.
Additionally, you can see from the logs that the mirror tries to synchronize the repository but fails many times in a row. It would be good if it could 'self-heal' on its own - perhaps doing a full resync, but understand this might not be so easy.