Open abc-moviestarplanet opened 8 years ago
in your VCS Root configuration use ssh
protocol git@github-enterprise:owner/repo.git
one more hack: with turned smudge
on checkout will be loooong)
so in your build agent configuration turn it off smudge = git-lfs smudge --skip %f
and create
command line buildstep with commands git lfs fetch && git lfs checkout
@msp-abc: It sounds like you're running github enterprise with SSH, but no SSL or HTTPS. That error message indicates it's trying to connect to an HTTPS port. Without HTTPS being enabled, that shouldn't work. So that indicates an issue with github enterprise. Can you send details to GitHub Enterprise support so I can see what version and config you're running so I can investigate? It doesn't sound like this is a git-lfs client issue.
You also mentioned setting up the lfs config in C:\Windows\System32\config\systemprofile\
. The git lfs install
command currently adds to your global git config, which is stored in your user home directory. This means it won't work for other users on the same system. We've talked in chat about changing this to default to the system config instead, with flags for specifying where you want the config explicitly (--system
, --global
, --local
).
@shuhrat: you can call git lfs pull
instead of fetch
+ checkout
:)
Thanks for the replies!
@shuhrat the VCS root configuration was set correctly. Why would smudge
on checkout take longer than doing a git pull
later? I will give it a try.
@technoweenie I have already sent a request to GitHub Enterprise support (request 26884). Does that suffice? Re: git lfs install
, yes, it would make sense to be able to specify whether to install it for the current or system user.
Why would
smudge
on checkout take longer than doing agit pull
later? I will give it a try.
Git calls smudge filters for each individual file in serial, which can be inefficient for repositories with lots of LFS objects. Going with @shuhrat's suggestion requires an extra git lfs pull
command, but this command is able to fetch the objects in parallel much faster.
I have already sent a request to GitHub Enterprise support (request 26884). Does that suffice?
Yup. Curious why no one pinged me on it :) I'll look into it.
@shuhrat i have tried implementing your workaround, but still find this error
Failed to perform checkout on agent: '"c:\program files\git\bin\git.exe" checkout -q -f master' command failed.
stderr: Downloading Assets/CustomAsset/Scripts/Editor/LabsterEditorTools/Resources/Gizmos/Green.png (4.73 KB)
Error accessing media: Assets/CustomAsset/Scripts/Editor/LabsterEditorTools/Resources/Gizmos/Green.png (731502e4ed000aae229aef2bcdeb3a6472d66b5ec7f2cee0d3f9256ceadf18c8)
Errors logged to .git\lfs\objects\logs\20160112T232528.7179394.log
Use `git lfs logs last` to view the log.
error: external filter git-lfs smudge %f failed 2
error: external filter git-lfs smudge %f failed
fatal: Assets/CustomAsset/Scripts/Editor/LabsterEditorTools/Resources/Gizmos/Green.png: smudge filter lfs failed
@adityayuwono in order to help you I should know what is in your configurations
.gitconfig
file on your agent?git lfs env
, add one command line step in Build Steps of your build configuration
just add commandssh
protocol? notice that ssh+git
won't workhey @adityayuwono -- aware I'm reviving a relatively old issue here, but just ran into the same issue at my work. inspecting the log on our agent we had a bad credential error; turns out someone had (inadvertently) removed the access token we were using for auth; I just had to regenerate the token and reset the VCS password configuration.
Hi all, I am going to close this thread, since it has been inactive for quite some time. I'm assuming that the underlying issue has been resolved or addressed in one of the many Git LFS releases since this was originally opened, but if this isn't the case, please do let me know and I would be happy to take a look into again sometime in the near future. Thanks!
I am having an issue trying to integrate Git LFS with TeamCity, using SSH key-based authentication.
Cloning and fetching works fine, but during checkout, it fails on the first LFS tracked file. I attached the LFS log here: git-lfs.txt
The point of interest:
Post https://<github-enterprise>/<organization>/<repo>.git/info/lfs/objects/batch: dial tcp 10.0.0.202:443: connectex: No connection could be made because the target machine actively refused it.
Now we do not have any SSL set up, so we want to use HTTP. When I try to clone the repository manually through cmd (on the TeamCity agent), everything works fine. The
git lfs env
also shows the same values, as seen here: git_lfs_env.txtI cannot see any difference except that the TeamCity agent is running under the Local System user.
I also had to create a .gitconfig file in the path
C:\Windows\System32\config\systemprofile\
for git LFS to automatically trigger during checkout (instead of running git lfs pull later in the build steps), with the following content:Any ideas what could be going wrong, or what else to check?
Currently I work around this by using HTTP instead of SSH, and hardcoding the username/password in the
.lfsconfig
in the repository but would like to avoid doing this.