Make it easier to delete local repository copies about rmt HOT 13 CLOSED

Here's how SMT implements this:

# smt-repos -h
   ...
   --delete                    : delete the repository from disk

from rmt.

Instead of actually implementing a "delete" command, we could print out where the repos are on the disk when the user calls repos disable or products disable. That gives users the chance to quickly remove the files if they want to.

from rmt.

#127 (comment) could be done as part of #94

from rmt.

@lagartoflojo I'm not sure if we want to do this with deduplication in place.

If a user deletes a directory that has a package used in symlinks, other products using that symlink could be broken. This would be fixed in the next mirror, but I'm not sure if we should find that "downtime" acceptable.

from rmt.

@tmuntaner but there are no symlinks? It's only hardlink or copy.

from rmt.

@ikapelyukhin You're right, you can ignore my previous comment ^_^

from rmt.

Hi guys, I'd like to take this one as @thutterer suggested me last time on #202 . Can somebody tell me where to start please ?

from rmt.

@NeimadTL At the moment there's no way to remove repositories that are no longer needed (repositories that are disabled for mirroring). Another issue is that repositories accumulate old packages over time: new versions of packages come out, repository metadata references new packages only -- which means that old packages become effectively inaccessible by client machines, but they are still stored on the disk of RMT server.

There are 3 possible things you could do:

1. The quick-and-dirty solution: when repository is being disabled, print the path to it on disk, so that the user can remove it manually. However, during mirroring all file names and checksums are stored in the DB (for creating hardlinks). This solution wouldn't address cleaning up the DB (that's the "dirty" bit).
2. Implement a new rmt-cli repos clean command that would go through the directories stored on the disk and remove the ones that aren't mirrored, as well as clean up entries in the DB.
3. Implement a new rmt-cli repos clean command that would go through the mirrored repositories and remove files that aren't referenced by the metadata anymore, as well as clean up entries in the DB.

That's the overview of things from my perspective. Please feel free to ask questions, you can also reach me on IRC, RocketChat or via email.

Thanks, Ivan.

from rmt.

Hi @ikapelyukhin, thank you for your guidance. I think I'll definitely have a lot more questions so I'll reach out to you one way or another tomorrow. No need to ask them here.

from rmt.

Hello @NeimadTL,

I just wanted to check in with you to see if you're still interested in working on this. If not, I'll create a task for the team to take over.

Thanks

from rmt.

I have to admit that I've put on the the side because I wanted to do too much things at the same time. I don't how a gem is organized in general and I don't know the RMT project in particular. Is that possible to give me one or two weeks to deal with it please ? if you have deadlines, I understand.

from rmt.

@lagartoflojo Do you think we can close this issue with the latest changes made to RMT?

from rmt.

Solved by #595

from rmt.