Capistrano error tar: This does not look like a ta

问题:

 INFO [050fe961] Running mkdir -p /home/rails/rails-
capistrano/releases/20140114234157 on staging-rails
DEBUG [050fe961] Command: cd /home/rails/rails-capistrano/repo && ( PATH=/opt/ruby/bin:$PATH GIT_ASKPASS=/bin/echo GIT_SSH=/tmp/rails/git-ssh.sh mkdir -p /home/rails/rails-capistrano/releases/20140114234157 )
 INFO [050fe961] Finished in 0.142 seconds with exit status 0 (successful).
 INFO [2dea2fe5] Running git archive feature/Capistrano | tar -x -C /home/rails/rails-capistrano/releases/20140114234157 on staging-rails
DEBUG [2dea2fe5] Command: cd /home/rails/rails-capistrano/repo && ( PATH=/opt/ruby/bin:$PATH GIT_ASKPASS=/bin/echo GIT_SSH=/tmp/rails/git-ssh.sh git archive feature/Capistrano | tar -x -C /home/rails/rails-capistrano/releases/20140114234157 )
DEBUG [2dea2fe5]    fatal: Not a valid object name
DEBUG [2dea2fe5]    tar: This does not look like a tar archive
DEBUG [2dea2fe5]    tar: 
DEBUG [2dea2fe5]    Exiting with failure status due to previous errors      

I am confused about two things:

1. Why is Capistrano running git archive here:
   git archive feature/Capistrano | tar -x -C /home/rails/rails-capistrano/releases/20140114234157

2. Why is tar failing?

回答1:

I had the same issue, until I realized I was pulling the nonexistent branch from git.

回答2:

Deleting app_name/repo also fixed this issue for me.

回答3:

This happens when the repo in the server to deploy is messed up. We're talking about the bare git repo that Capistrano by default would put in /var/www/$application/repo (for other people's reference).

In your case it does not have a local feature/Capistrano branch so when running git archive feature/Capistrano nothing is output to that | pipe. To confirm, ssh into the server, cd into /home/rails/rails-capistrano/repo, and run git branch.

1. It's running git archive as a way to export the selected branch's tree. git archive "writes it out to the standard output" so Capistrano redirects that to tar in order to uncompress the archive immediately into your new release directory. (Why Capistrano chose this instead of git checkout defeats me.)

2. tar fails because it's receiving nothing d:

I can think of two possible solutions/ways to troubleshoot:

• ssh into the server and manually delete the repo folder (e.g. in your case /home/rails/rails-capistrano/repo) as mentioned by @lugolabs
• make sure the server's repo is using the remote you're expecting (ssh in, cd into repo/, and run git remote -v) -- you may just need to update your :repo_url in deploy.rb (and delete the repo/ dir).

回答4:

I think that folder gets populated via a git pull, so it shouldn't be empty. If you do see it empty the issue is from the git not the tarball.

The issue I had was my capistrano deploy.rb repository URL was set to a different one than that project i was working in. In order to fix this issue, I also had to logon to the server and delete the app_name/repo folder which must have been caching the original bad remote URL.

回答5:

Whenever I have hit this error it was because the branch specified in my deploy/environment.rb file wasn't checked into git. Do an add / commit / git push origin branch_name and that will likely make things work.

回答6:

I'm using Bedrock Roots (wordpress) for development, capistrano for deploys and git flow. Stumbled upon this error when tried to deploy, while on hotfix/x.x.x branch locally. So I finished current (merged changes to develop branch) and then successfully deployed.