Merge branch 'master' of ssh://git-annex.branchable.com

This commit is contained in:
Joey Hess 2020-10-27 14:22:21 -04:00
commit 970e4f11e9
No known key found for this signature in database
GPG key ID: DB12DB0FF05F8F38
3 changed files with 30 additions and 0 deletions

View file

@ -0,0 +1,13 @@
I noticed today when running `git annex sync` that it was unhappy with one of my remotes:
```
remote: git-annex: Repository /home/anovak/annex/.git is at unsupported version 7. Automatic upgrade exception! actually-broken.log: getSymbolicLinkStatus: does not exist (No such file or directory)
```
Digging in, it looks like that machine is running Ubuntu 20.04, which ships Git Annex 8, which *only* supports repo version 8. The rest of my machines are a mix of Ubuntu 18.04 and other Debian derivatives, which all ship older versions of Git Annex, which *can't* support repo version 8.
Is there a way for me to get out of manually installing Git Annex on every machine to upgrade everything to v8 at the same time?
And, before I've upgraded, is there a way to screw up my repo by e.g. running `git annex sync` from the newer Git Annex and pushing commits in the new format over to machines that still only speak the old format?
How are other people handling this transition where there's no repo format that's speakable by all the machines? Just upgrading everything at once and replacing the distro's git annex with a manually installed version?

View file

@ -0,0 +1,8 @@
[[!comment format=mdwn
username="Lukey"
avatar="http://cdn.libravatar.org/avatar/c7c08e2efd29c692cc017c4a4ca3406b"
subject="comment 1"
date="2020-10-27T09:06:38Z"
content="""
I just use the (daily) standalone linux build everywhere, not the package from my distro. I extract it to `/opt`, set `GIT_ANNEX_PACKAGE_INSTALL=no` at the top of `git-annex.linux/runshell` and then link git-annex to `/usr/bin/git-annex` with `ln -s /opt/git-annex.linux/git-annex /usr/bin/git-annex`.
"""]]

View file

@ -0,0 +1,9 @@
[[!comment format=mdwn
username="michael.hanke@c60e12358aa3fc6060531bdead1f530ac4d582ec"
nickname="michael.hanke"
avatar="http://cdn.libravatar.org/avatar/f881df265a423e4f24eff27c623148fd"
subject="Documentation of demand"
date="2020-10-27T14:59:47Z"
content="""
We are routinely working with datasets in the 100TB range, with individual file sizes of up to 16TB. Some of our systems have ~500MB/s disk read throughput only. The separation of download and checksuming leads to 6+ hours of additional disk reads and delay the availability of a file. Having this feature would be a stellar performance improvement for our use case. Thanks in advance.
"""]]