After moving some files (about 1G, some big and some small files) in and out the annex, I noticed that the size of the repository has grown quite a bit. My empty repository now is over 100 MB (even after "git annex dropunused ..." and "git gc").
Most of this size is not Git metadata but many small files in the git-annex branch that seem to hold information about files I deleted (even in the other known repositories).
So is there a way to get rid of these useless but space consuming information?
(Maybe there is no (elegant) way to remove the symlink versions from the Git history (is there?), but it would already be nice if those small git-annex metadata files could be removed)
You must have quite a number of files and/or location tracking churn to get the branch that large.
Removing location tracking files for vanished files from the git-annex branch would not save much space.. Only a very little bit of tree object size going forward. The way the git-annex branch is merged does not really allow deleting files from it. And of course git uses space for even deleted files.
It'd be possible to delete the git-annex branch entirely, first backing up git-annex:uuid.log git-annex:remote.log and git-annex:trust.log and then making a new, clean branch that those files are added back to. Then run git annex fsck on every clone to repopulate the location tracking info for files that still exit. (Note that this would also lose urls stored by
git-annex addurl.)