Migrating Repository without Network

I have a server with a git repository whose NIC is no longer working. I need to migrate that repository to a new server using a flash drive. I can only find instructions for migrating when both servers have network access. Can someone tell me how to do a migration without using a network? TIA.

On Tue, Oct 21, 2025 at 10:52 AM Dennis Putnam <dap1@bellsouth.net> wrote:

Have a look at git-bundle(1), which was designed for something similar to this use case I believe. Or, rsync the repo to the flash drive and then off?

"D. Ben Knoble" <ben.knoble@gmail.com> writes:

But bundle is only about preserving/transporting the history recorded in the repository. It does not do the repository config or the hooks. Those folks would want to migrate the entire repository data off of a failing server, so I think "rsync/cp" would be the more sensible route.

Thanks.

On Tue, Oct 21, 2025 at 10:21:52AM -0400, Dennis Putnam wrote:

Note that Git repositories - both "bare" which are typical for servers [1] - and "normal" - which usually get created when the folks clone repositories from centralized locations, are self-containing, so what you should do is to just physcally copy the directory with the repository to the flash drive, recursively. Then make sure you have another copy of it, just in case you'll screw something later.

Then, once you have the repository on the flash drive, what to do heavily depends on how the access to that repository was provided - simply because there are many ways to do that - say, via SSH with several real system accounts, via SSH with a single "virtual" user, via HTTP[S] with the authentication provided by the mediating web server, and so on; it is also possible to combine these ways of access. It is also possible that the repo is under the control of some turn-key solution such as GitLab, Gitea etc. So, in the end, what to do next depends on what setup *around* that Git repo you have right now.

In either case, it will be possible to fetch all the data from the repo saved on the flash drive: it might be not exactly obvious but all Git commands which fetch data - such as "clone", "fetch" etc - do understand "plain" pathnames such as /the/path/to/the/repo and URLs with the special scheme "file://" - such as file:///the/path/to/the/repo, allowing them to work with on-disk repos available locally such as on a mounted flash drive.

 1. https://git-scm.com/book/en/v2/Git-on-the-Server-Getting-Git-on-a-Server

On Wed, Oct 22, 2025 at 12:04:15PM +0300, Konstantin Khomoutov wrote:

[...]

...it should be noted, though, that "normal" and "bare" repository have different schemes of storing branches available in them, and it might affect the approach you will need to employ to fetch really *all* the data to form a new repo.

It's hardly possible to dive further in this topic because we have next to zero knowledge about your setup - including whether that repo is bare and is accessed from the outside, or normal and merely shared between the users physically logged into the server and so on. I mean, it may so happen that all you will need to do is to copy the repo to a flash drive and then copy it again to the new server, possibly fix filesystem permissions on it so that they match the ones the old repo has, and call it a day. In case you will need more advanced help, you'll need to provide more details on your situation.

Thanks for the reply. I'm not a git expert (obviously) so I am not sure of some of the terminology. This is a remote repository that is accessed via SSH from various clients. Does that clarify things?

On 10/22/2025 5:04 AM, Konstantin Khomoutov wrote: