Question

Currently doing some R&D about file synchronization, I eventually stumbled upon rsync (who wouldn’t ?).

Context: There’s a server used as central file repository, that is updated only by an admin. Several users synchronize their filesystem from the central repo. It’s one-way sync, meaning that users will never update the repository. The updates are made with rsync.

Let’s consider a little scenario:

1. User starts updating files from the server and downloads file A
2. Admin updates the files A and B on the server, creating A’ and B’

Question: What will be the third step ?

• User downloads B’. He has A and B’, resulting in an inconsistent state among files
• User downloads B. He has A and B, resulting in a consistent but outdated state
• User downloads B’ and A’, resulting in a consistent and up-to-date state
• Anything else ?

Answer 1

As far as I know, it’s your first option: User will have A and B’. The only way to avoid this is to snapshot your filesystem before the rsync, but somehow this feels like massive overkill.

Other solutions might be

• put A and B together in one zip/tar file (the administrator/the one updating, has to do this; else files can still be updated between adding the first and second file
• Use git (or another VCS) as central repository