This plugin for Capistrano 3 will add a timestamped Git tag at each deployment, automatically. It requires :branch and :stage to be set, but as Capistrano 3 is multistage by default (unlike Cap 2) :stage should already be set, but you can override the variable if you want to change the name of the tag.
As of version 1.0.7, this plugin requires Cap 3.7.
If you need a Capistrano < 3.7 compatible version, then use gem 'capistrano-deploytags', '1.0.6'
If you need a Capistrano 2 compatible version, then use gem 'capistrano-deploytags', '~> 0.9.2'
Simply: it makes it so you can track your deployments from Git. If I were to issue the command:
cap production deploy
This would result in one new git tag with the environment and timestamp:
production-2012.04.02-203155-utc
These tags can be used for any number of useful things including generating statistics about deployments per day/week/year, tracking code size over a period of time, detecting Rails migrations, and probably a thousand other things I haven't thought of.
capistrano-deploytags is available on
rubygems.org.
In keeping with the pattern used by Capistrano itself and other plugins, add it
to the development
group of your Gemfile with require: false
:
# Gemfile
group :development do
gem 'capistrano-deploytags', '~> 1.0.0', require: false
end
Then require capistrano/deploytags
in your Capfile
# Capfile
require 'capistrano/deploytags'
This will create two tasks, one that runs before the deploy
task, and one
that runs after the cleanup
task.
NOTE: You will be creating and pushing tags from the version of the code in the current checkout. This plugin needs to be run from a clean checkout of your codebase. You should be deploying from a clean checkout anyway, so in most cases this is not a restriction on how you already do things. The plugin will check if your code is clean and complain if it is not.
ALSO: The plugin will do a pull to make sure you have the code on your local system that will actually be deployed before checking the tree for changes. Know this ahead of time as this may affect how you deal with your deployment branches.
By default, Capistrano Deploytags will use the remote names origin
. If you
use a different remote name, then you may change the :git_remote
setting
from your deploy.rb
or the stage.
Because you must have a clean tree to deploy, working on your deployment scripts themselves can be a bit frustrating unless you know how to make it work. The easiest way around this problem is to simply commit your changes before you deploy. You do not have to push them. The plugin will then happily carry on deploying without complaint.
Alternatively, you could disable the plugin temporarily with one of the methods described below.
Sometimes you do not want to enable deployment tagging for a particular
stage. In that event, you can simply disable tagging by setting no_deploytags
like so:
set :no_deploytags, true
You can also set this from the command line at any time with an environment
variable cap stage deploy NO_DEPLOYTAGS=true
.
NOTE: this will disable the use of the plugin's functionality entirely for that stage. The tasks will run, but will do nothing. This means that tasks that are hooked to the Capistrano Deploytags tasks will also still run, but they may find their expectations are not met with regards to the cleanliness of the git tree.
You may override the time format in deploy.rb
or your stage:
set :deploytag_time_format, "%Y.%m.%d-%H%M%S-utc"
To use your local time and not UTC (so that Time.now
and not Time.now.utc
is used internally):
set :deploytag_utc, false
By default, Capistrano Deploytags will create a tag with a message that indicates
the local user name on the box where the deployment is done, and the hash of the
tagged commit. If you prefer to have a more detailed commit message you may override
the :deploytag_commit_message
setting from your deploy.rb
, e.g.
set :deploytag_commit_message, 'This is my commit message for the deployed tag'
It's trivial to view the deployment history for a repo. From a checkout
of the repo, type git tag -l -n1
. The output looks something like:
dev-2013.07.22-105130 baz deployed a4d522d9d to dev
dev-2013.07.22-113207 karl deployed 4c43f8464 to dev
dev-2013.07.22-114437 gavin deployed 776e15414 to dev
dev-2013.07.22-115103 karl deployed 619ff5724 to dev
dev-2013.07.22-144121 josh deployed cf1ed1a02 to dev
A little use of grep
and you can easily get the history for a
particular (e.g. git tag -l -n1 | grep dev
).
It should be noted that the names used when tags are created are the local user name on the box where the deployment is done.
You might find it useful to add this to your ~/.gitconfig in order to get a nice history view of the commits and tags.
[alias]
lol = log --pretty=oneline --abbrev-commit --graph --decorate
You can then view the list by typing git lol
from the checked out
code path.
Because you have to actually be on the head of the branch you are deploying in order for tagging to work properly, deploying a previous commit doesn't work as you might expect.
One simple solution is to configure your config.rb
to accept an ENV var
override. Then if you need to deploy a previous commit you can check out that
commit (SHA or branch), and supply the var on the command line. e.g. with this
in your config.rb
:
set :branch, ENV["REVISION"] || ENV["BRANCH_NAME"] || "master"
you can deploy a previous commit with
git checkout <previous-commit>
cap <stage> deploy REVISION=<previous-commit>
Because Jenkins will check out the code with the current revision
number you will be in a detached state. This causes the plugin to be
unhappy about the git tree. The solution is to add -S branch=$GIT_COMMIT
to the cap deploy line called from your Jenkins build. This will cause
the diffs and comparisons done by the deploytags gem to be correct.
This software was written by Karl Matthias with help from Gavin Heavyside and the support of MyDrive Solutions Limited.
This plugin is released under the BSD two clause license which is available in both the Ruby Gem and the source repository.