Manually creating a lock

[sorcix]

Output of restic version

restic 0.9.4 compiled with go1.11.4 on darwin/amd64

What should restic do differently? Which functionality do you think we should add?

Add the ability to manually create a write lock.

What are you trying to do?

My restic repository is stored both in the cloud (master) and on a USB disk drive (backup copy). For performance reasons I run the check/prune commands on the USB drive, then upload the changes using rclone. This means that while I'm running check/prune the cloud copy is not locked. It would be great if there was a way to manually create a lock.

[sorcix]

Looks like most of the code is already there in the cmd/restic package and I just need to create a new command to make use of it. If it's ok to have this feature I could create a PR to add it.

[ProactiveServices]

I'd definitely make use of this in a similar use-case.

[MichaelEischer]

If we implement this, I think the command should be modeled after borg [common options] with-lock [options] REPOSITORY COMMAND [ARGS...]. Simply creating a lock file won't be sufficient as it must be refreshed from time to time.

[alyandon]

I have a similar use case for a feature like this. I'm in the process of migrating from borg to restic and I need to sync my repos to off-site storage for DR purposes and would like to protect against restic updating the repo while the sync process is running. Currently, with borg - the sync process can be run via borg with-lock <repo> <command> which locks the repository for exclusive access for the duration of the operation.

[MichaelEischer]

restic never modifies a file after writing it (the full lifecycle is: write temp file, rename to final name, read n times, delete). You can copy a repository while restic is active, by first syncing the snapshots directory, then the index and finally data (first copy new files, then remove deleted ones). If the sync process does not complete, the some new snapshots may not be readable, however, existing snapshots won't be affected. (It might be necessary to run restic repair index though)

[alyandon]

That is certainly good information to know!

I've already cobbled together a somewhat working solution via flock for the time being but that also involves writing a wrapper script to check for the lock (e.g. wrapper script named restic - actual restic binary as restic-bin) to get everything working.