- If/when I'm employing anyone else then things are taking off and CI will be a good idea but likely to be left behind in distraction;
- It keeps me honest, by making sure there's no silly dependencies to my laptop and running stats on the code; and
- It keeps my hand in with sensible tools. In a similar vein I use Assembla for managing tasks and as a git repository.
AWS ConfigFirst, I brought up an AWS EC2 instance. I chose the Amazon 32bit model. Why 32bit? - because I want to run this in a t1.micro (free) instance. I added an EC2 disk, of 2GB, to hold the builds and Jenkins data, that might want to be persistent; and gave it an imaginative name.
Linux ConfigUsing the AWS AMI the usual install and post-install tweaking are much reduced needed, which is nice. The steps were:
Use the key generated above to login with ssh, do the initial update and mount the extra disk:
dan-laptop$ cp ~/Downloads/awskey.pem ~/.ssh/ dan-laptop$ cd .ssh/ dan-laptop$ chmod 600 awskey.pem dan-laptop$ ssh -i ~/.ssh/awskey.pem email@example.com __| __|_ ) _| ( / Amazon Linux AMI ___|\___|___| https://aws.amazon.com/amazon-linux-ami/2013.03-release-notes/ There are 5 security update(s) out of 10 total update(s) available Run "sudo yum update" to apply all updates. [ec2-user ~]$ sudo yum update Loaded plugins: priorities, security, update-motd, upgrade-helper amzn-main | 2.1 kB 00:00 amzn-updates | 2.3 kB 00:00 Setting up Update Process Resolving Dependencies --> Running transaction check ---> Package aws-apitools-ec2.noarch 0:220.127.116.11-1.0.amzn1 will be updated ... [ec2-user ~]$ df Filesystem 1K-blocks Used Available Use% Mounted on /dev/xvda1 8256952 1059968 7113124 13% / tmpfs 307704 0 307704 0% /dev/shm [ec2-user etc]$ sudo mkfs -t ext4 /dev/sdb mke2fs 1.42.3 (14-May-2012) Filesystem label= OS type: Linux Block size=4096 (log=2) Fragment size=4096 (log=2) Stride=0 blocks, Stripe width=0 blocks 131072 inodes, 524288 blocks 26214 blocks (5.00%) reserved for the super user First data block=0 Maximum filesystem blocks=536870912 16 block groups 32768 blocks per group, 32768 fragments per group 8192 inodes per group Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912 Allocating group tables: done Writing inode tables: done Creating journal (16384 blocks): done Writing superblocks and filesystem accounting information: done [ec2-user ~]$ cd /etc [ec2-user etc]$ sudo cp fstab fstab.install [ec2-user etc]$ sudo vi fstab [ec2-user ci]$ diff /etc/fstab.install /etc/fstab 2a3 > /dev/sdb /data/ci ext4 defaults 1 2 [ec2-user etc]$ sudo mount /dev/sdb /data/ci [ec2-user etc]$ cd /data/ci [ec2-user etc]$ sudo ln -sf /usr/share/zoneinfo/Europe/London /etc/localtime
Install various packages that we need to integrate with, and accepting all dependancies:
sudo yum install erlang wget http://www.rabbitmq.com/releases/rabbitmq-server/v3.1.3/rabbitmq-server-3.1.3-1.noarch.rpm sudo yum install rabbitmq-server-3.1.3-1.noarch.rpm sudo chkconfig --levels 235 rabbitmq-server on sudo service rabbitmq-server start sudo yum install mysql mysql-server mysql-libs sudo service mysqld start /usr/bin/mysqladmin -u root password 'new-password' sudo chkconfig --levels 235 mysqld on sudo yum install mysql-connector-java cd /data/ci sudo mkdir mysql sudo /etc/init.d/mysqld stop sudo cp -R -p /var/lib/mysql/ /data/ci/mysql/ sudo cp /etc/my.cnf /etc/my.cnf.install sudo vi /etc/my.cnf diff /etc/my.cnf /etc/my.cnf.install 2c2 < datadir=/data/ci/mysql --- > datadir=/var/lib/mysql sudo /etc/init.d/mysqld start cd /etc/yum.repos.d sudo vi 10gen.repo cat /etc/yum.repos.d/10gen.repo [10gen] name=10gen Repository baseurl=http://downloads-distro.mongodb.org/repo/redhat/os/i686 gpgcheck=0 enabled=1 sudo yum install mongo-10gen mongo-10gen-server sudo chkconfig --levels 235 mongod on sudo service mongod start yum search java | grep 'java-' sudo yum install java-1.7.0-openjdk.i686 sudo alternatives --config java wget http://downloads.typesafe.com/play/2.1.2/play-2.1.2.zip cd /usr/local unzip ~/play-2.1.2.zip sudo chown -R root:root play-2.1.2/ yum search tomcat | grep tomcat7 sudo yum install tomcat7-lib.noarch tomcat7-servlet-3.0-api.noarch tomcat7.noarch sudo yum install ant sudo yum install git
To support selenium for web testing I installed Xvfb and firefox, to run headless. Firefox doesn't install on an AWS AMI so easily, but the script at Joe Killer's blog seems to do the job after a couple of hours of compiling stuff (although it doesn't like a bz2 compression at the end and for my 32 bit install the URL to get firefox from at the end needs changing to http://releases.mozilla.org/pub/mozilla.org/firefox/releases/latest/linux-i686/en-GB/). Xvfb is then
sudo yum install Xvfb Xvfb :99 -screen 0 800x600x16&The starting of X can also be done in the Jenkins job later, but doesn't seem to exit cleanly.
JenkinsI'm using Jenkins as a continuous integration server. It seems to do things I want of it - I haven't been moved to try the alternatives yet.
sudo wget -O /etc/yum.repos.d/jenkins.repo http://pkg.jenkins-ci.org/redhat/jenkins.repo sudo rpm --import http://pkg.jenkins-ci.org/redhat/jenkins-ci.org.key sudo yum install jenkinscd /data/ci sudo mkdir jenkins sudo chown -R jenkins:jenkins jenkins sudo cp /etc/sysconfig/jenkins /etc/sysconfig/jenkins.install sudo vi /etc/sysconfig/jenkins sudo diff /etc/sysconfig/jenkins /etc/sysconfig/jenkins.install 10c10 < JENKINS_HOME="/data/ci/jenkins/" --- > JENKINS_HOME="/var/lib/jenkins" 80c80 < JENKINS_HANDLER_MAX="20" --- > JENKINS_HANDLER_MAX="100" 88c88 < JENKINS_HANDLER_IDLE="10" --- > JENKINS_HANDLER_IDLE="20" sudo /etc/init.d/jenkins start sudo chkconfig --levels 235 jenkins on
Then the Jenkins config can happen via the web interface. Manage Jenkins is the start point, where it will encourage securing the install. Take care not to lock yourself out: switch on security, but let users register and do anything. Then create an admin user, and then close registrations!
Add plugins, git and assembla auth for now:
AssemblaI use Assembla for task organisation and as a git repository. The obvious thing to do is to get Jenkins to get code from git when I push a change. I have an integration branch in git just for this. (Most) development happens in a branch off integration and gets merged back in to integration when it passes unit tests. Live deployment, when it happens, will be by merging from integration into a main branch when integration (and happy testing) pass.
For Jenkins to use git get updates from Assembla we need a key with an empty passphrase, which will just be given read permission. The key in ~jenkins/.ssh/id_rsa.pub is then added to Assembla's keys. The sudo trick gets round the fact that the jenkins user isn't designed to be logged-in.
sudo su - -s /bin/bash jenkins ssh-keygen -t rsa -C "jenkins@aws" git ls-remote -h firstname.lastname@example.org:projectname.git HEAD
Now, the key needs to be registered with Assembla:
Finally, assembla can be swapped in as an authorisation mechanism. The assembla auth plugin allows the assembla user name to be used to log into Jenkins (not to be the git key). This saves managing another set of user ids. As for the ssh key, this is a step that needs changes on both sides. Additional users are added by creating their own API key and adding them to the list in the Jenkins security page.
Build for CIFor me one of the key benefits of CI is that you're regularly building on a not-a-development-system. Ideally on something somewhat like a deployment system. This means that build scripts, like code, need to have any system specific stuff (paths, ports etc) grouped into a few config files. For me that means ant and properties files. The first couple of builds on the new CI server will fail while you tweak out database users that don't exist, paths that have changed, correct passwords etc. Build config may get its own post sometime later.
- Get jobs to stop and start their own services (mysql, mongo, xvfb, rabbitmq, tomcat...) as a micro instance doesn't have enough memory for all this lot at once! Indeed, this will probably need a slave with more memory, on demand, for some tests.
- Make an AMI from the system and take a snapshot of the /data/ci filesystem, to speed up reuse.
- Metrics, graphs, alerts...
- Test deployment on the CI box
- Tests which exercise the system via http.