I would like to discover how much data is stored in a special remote across repositories. But it might take annex many minutes to figure out other stats like size of the files in the tree etc, which I do not care. So I wondered if for e.g.
(venv-annex) dandi@drogon:/mnt/backup/dandi/dandisets$ git -C 000003 annex info --in dandi-dandisets-dropbox
trusted repositories: 0
semitrusted repositories: 3
00000000-0000-0000-0000-000000000001 -- web
00000000-0000-0000-0000-000000000002 -- bittorrent
b7fcf214-e492-4f2c-8789-708af9fd4656 -- dandi@drogon:/mnt/backup/dandi/dandisets/000003 [here]
untrusted repositories: 1
727f466f-60c3-4778-90b2-b2332856c2f8 -- [dandi-dandisets-dropbox]
transfers in progress: none
available local disk space: 24.88 terabytes (+100 megabytes reserved)
local annex keys: 101
local annex size: 2.56 terabytes
annexed files in working tree: 101
size of annexed files in working tree: 2.56 terabytes
combined annex size of all repositories: 7.68 terabytes
annex sizes of repositories:
2.56 TB: 00000000-0000-0000-0000-000000000001 -- web
2.56 TB: 727f466f-60c3-4778-90b2-b2332856c2f8 -- [dandi-dandisets-dropbox]
2.56 TB: b7fcf214-e492-4f2c-8789-708af9fd4656 -- dandi@drogon:/mnt/backup/dandi/dandisets/000003 [here]
backend usage:
SHA256E: 101
bloom filter size: 32 mebibytes (0% full)
I could just (quickly) get
untrusted repositories:
727f466f-60c3-4778-90b2-b2332856c2f8 -- [dandi-dandisets-dropbox]
annex sizes of repositories:
2.56 TB: 727f466f-60c3-4778-90b2-b2332856c2f8 -- [dandi-dandisets-dropbox]
with --json also correspondingly trimmed up. Or it could potentially be a different record output entirely, concentrating on that remote?
The "annex sizes of repositories" table is indeed what you want. Since about a year ago, git-annex maintains a running total of the sizes of all repositories. So it can generally get that information very fast.
In cases where it needs to do work to update the running total, it has to replay changes to the location log, which is the expensive bit. Updating the running totals for all repositories does not really impact the speed. So focusing on the size of a specific remote doesn't seem useful.
(Using
--into do it would also overload the meaning of that option in a confusing way, bearing in mind that it can already be used with a command likegit-annex info --in=here .)I think that what is needed is a way to make git-annex info only generate specific parts that you request, and skip the work to calculate other parts. Eg:
The --fast option kind of does this, for things that can be generated really quickly. The annex sizes of repositories does not quite fit in --fast though, since it could take a long time in some edge cases to update the running totals.
I think this --show option would be easy to add to info.
info --show