How to deploy from Bamboo using Git?

Question

How can I deploy from OnDemand Bamboo to an external server via a Git repo?

What I had envisioned was:

* Have Bamboo checkout a branch from the Git repo (on Bitbucket if that matters).
* Run targets from a build.xml file stored in the repo, these targets would run Unit Tests or minimize CSS and JS files. In other words, perform some actions on the files.
* On build success, create a zip artifact or create a new tag and commit that to the repo.
* Have Bamboo connect, maybe via SSH, to the web host and execute a Git pull to update the code on the server.

The problem seems to be that, because I'm using Git, I can't push back to the repo as I want to do in #3. Then, I don't have a way to connect to the web host to execute the pull because I can't seem to automate an SSH call that uses user/pass credentials.

Am I thinking about this wrong? Is this possible in a downloaded instance but not in OnDemand? Am I missing something obvious?

josephsong · Answer

I think I follow most of that, but what I realized is I can do almost all of it through running scripts as a task in the build. What I'm still stuck on, however, is connecting to the webhost server via SSH. The only way I could figure out to do it was to install the SSH plugin from Atlassian Labs, but of course you can't install a plugin in the OD instance.

Maybe I should put it this way: how can I run commands on a remote server from a OD Bamboo task?

Przemek Bruski · Answer

Joe, there shouldn't be any difference.

"Push tag" part

On OnDemand, the checkout will be made from HTTP url, so no credentials will be available to the agent and you need to make sure they're there.

Also, like all checkouts made against a specific revision, your checkout may be a 'detached head' checkout, meaning that you need to make sure you're repo is appropriately synced with origin so that the tag/push call suceeds.

Finally, IIRC, the repository will have no origin defined, so you need to specify the push destination.

"Pull" part

If you use HTTPs access for pull, you can inject user/pass directly into the URL - this is the simplest method.

Otherwise, you need to make sure you SSH has access to credentials you want to use.