Automating build installation process - linux

Question

I work on a SaaS based product of my company which is hosted on private cloud. So every time a fresh BOM package is available by the DEV team. In the common share folder , we- the testing team installs the build on our application servers (3 multi node servers, with one being primary and the other two being secondary).
The build installation is entirely done manually.on the three app servers(linux machine), where in the steps we follow are as below
Stop all the app servers
Copy the latest build from a code repository server(copy the .zip build file)
Unzip the content s if the folder on to a folder in the appserver (using the unzip command)
Run backup of existing running build on all three folders( command is something like - ant-f primaryBackup.xml, ant-f secondary backup.xml )
Then run the install on all three serverscommand is something like - ant-f primaryInstall.xml, ant-f secondaryInstall.xml )
Then restart all the server and check if the latest build is successfully applied.
Question: I am wanting to automate this entire process, such that I am just required to give the latest build number to be installed and the script takes care of the whole installation .
Presently I don't understand how this can be done ? Where should I start? Is this feasible? Will a shell script of the entire process be the solution?

Answer 1

There are many build automation/continuous deployment tools out there that would help you with a solution for automating your deployment pipeline. Some of the more popular configuration automation tools out there are puppet, chef, ansible, and saltstack. I only have experience with ansible and chef but my interpretation has been that chef is the more "user-friendly" option. I would start there... (chef uses the ruby language and ansible uses python).
I can answer specific questions about this, but hour original question is really open ended and broad.
free tutorials: https://learn.chef.io/
EDIT: I do not suggest provisioning your servers/deployments using bash scripts... that is generally messy and as your automation grows (which it likely will), your code will gradually become unmanageable. Using something like chef, you could set periodic checks for new code in your repositories and deploy when new code is detected (or upon certain conditions being met). you could write strait bash code within a ruby bock that will remotely stop/start a service like this (example):
bash 'Copying the conf file' do
cwd "current/working/directory"
user 'user_name'
code <<-EOH
nohup ./startservice.sh &
sleep 2m
nohup ./startservice.sh &
sleep 3m
EOH
end
to copy code from git for example... I am assuming github in this example, as i do not know where your code resides:
git "/opt/mysources/couch" do
repository "git://git.apache.org/couchdb.git"
reference "master"
action :sync
ssh_wrapper "/some/path/git_wrapper.sh"
end
lets say that your code is anywhere else.. bamboo or Jenkins for example... there is a ruby/chef resource for it or some way to call it using strait ruby code.
This is something that "you" and your team will have to figure out a strategy for.
You could untar a file with a tar resource like so:
tar_package 'http://pgfoundry.org/frs/download.php/1446/pgpool-3.4.1.tar.gz' do
prefix '/usr/local'
creates '/usr/local/bin/pgpool'
end
or use the generic linux command to like so:
execute 'extract_some_tar' do
command 'tar xzvf somefile.tar.gz'
cwd '/directory/of/tar/here'
not_if { File.exists?("/file/contained/in/tar/here") }
end
You can start up the servers in the way that I wrote the first block of code (assuming they are services.. if you need to restart the actual servers, then you can just run init 6 or something.
This is just en example of the flexibility these utilities offer