previous From the ground up
There are any number of blog posts advertising some pretty wacky ways to set up your computer to use different identities for the same host[1] depending on the repository you’re working in. Most involve configuring SSH to listen out for the hostname mangled with the name of the identity to be used which is then unmangled in the ~/.ssh/config file. I didn’t find this to be a satisfactory solution and never used it.
There are other ways!
Since Git version 2.3.0 (Feb 2015), an environment variable GIT_SSH_COMMAND has been read by the program to replace, as you might imagine, the SSH command used when talking over the network using SSH.
For some time I used this to “manage” identities when interacting with GitHub remotes, but it wasn’t perfect and and ended up with eye-wateringly ugly commands like:
GIT_SSH_COMMAND='ssh -i ~/.ssh/id_rsa_job_a’ git push origin ben/local-name:ben/remote-name
which are hard to read. And hard to read is upsetting.
Since Git version 2.10.0 (Aug 2016), it has been possible to access the equivalent functionality offered by GIT_SSH_COMMAND environment variable through Git’s config file. The new config variable is core.sshCommand and (as it’s so concisely stated in the release notes) it allows the user to specify what value for GIT_SSH_COMMAND to use per repository.
Just set core.sshCommand as any other Git config variable, specifying the identity that is to be used for that repository:
git config core.sshCommand 'ssh -i ~/.ssh/id_rsa_job_a'
and next time you run a SSH-requiring Git command from that repo, the specified identity will be used.
| [1] | Read github.com |
Serverless PyPI next