A ZF STORY:

Parallel made easy

Gianluca Arbezzano / @ CurrencyFair

Gianluca Arbezzano

Software Engineer at CurrencyFair

OpenSource maintainer

twitter.com/gianarb - github.com/gianarb

Docker Captain and DevOps culture's follower

twitter.com/gianarb - github.com/gianarb

Zend Framework 3

Make order inside our dependencies

Mvc works without Log, Cache...

Small components with single responsability

Mvc redesign and split of FlashMessage

Decouple Ldab and ACL

Defence from bad implementation

bye bye AwareInterface

Parallel computing is a type of computation in which many calculations are carried out simultaneously, operating on the principle that large problems can often be divided into smaller ones, which are then solved at the same time. by. wikipedia

Image copyleft Hyde & Rugg, 2014, incorporating house icon from Wikimedia

Image copyleft Hyde & Rugg, 2014, incorporating house icon from Wikimedia

the builder must check to have enough wheelbarrows, trowels and people

For us it is easier because we have cloud computing.

Put another way, a five-person company can harvest 10,000 servers with no infrastructure investment, all at the push of a button. That can be pretty disruptive to industries where computing power is everything.

GreenButton ex. CTO Dave Fellows

Case Study

We worked to split zendframework/zf2 repository and its history in different and independent repositories

Why

  • maintainability
  • component first
  • Remove old split flow based on git subtree

Numbers

  • 6 years of git's history
  • ~27k commits, 67 releases, and over 700 contributors
  • clean checkout is around 150MB

by mwop's blogpost

Results

50 repositories

Zend\Http component repository ended up with ~1.7k commits, 50 releases, ~160 contributors, and a clean checkout clocks in at 5.4MB

by mwop's blogpost

time

Between 5-12 hours

for each repository

via GIPHY

Obvious! Each component could be split easy in parallel because the process doesn't require nothing.
It can leave alone

Made easy because after one email AWS allowed our account to power on 55 EC2 instances

We are worked on two different directions

  • A procedure to manipulate the git's history
  • A provisioning flow for our servers

Provision tools

We have a specific and unique use case.
PHP and the AWS sdk could work!!

ZF Parallel Split

github.com/gianarb/zf-parallel-split

This project contains the procedure to boot one server for component and inject the init boot script


return [
    "aws" => [
        'key' => '',
        'secret' => '',
        'region' => "us-east-1",
    ],
    "githubToken" => "",
    "iam" => [
        "name" =>,
    ],
    "keyName" => "gianarb-def",
    "bucketBackup" => "zend-split",
    "sshPath" => "ssh/",
    "components" => [
        "Authentication",
        "Barcode",
        "Cache",
        "...",
    ]
];
                    

$result = $client->runInstances(array(
    "ImageId" => "ami-d05e75b8",
    "MinCount" => 1,
    "MaxCount" => 1,
    "InstanceType" => "t2.medium",
    "InstanceInitiatedShutdownBehavior" => "terminate",
    "KeyName" => $config['keyName'],
    "UserData" => base64_encode(render($config, $component)),
    'IamInstanceProfile' => array(
        'Name' => $config['iam']['name'],
    ),
));
                    

Launch a t2.medium for each component and create a specific UserData script


#!/bin/bash
RAM_DISK=3000m
mount -t tmpfs -o size=$RAM_DISK tmpfs /root
add-apt-repository -y ppa:git-core/ppa
apt-get update
apt-get install -y python2.7 curl git php5 php5-curl php5-cli
git config --global user.name "Gianluca Arbezzano"
git config --global user.email [email protected]
curl "https://bootstrap.pypa.io/get-pip.py" -o "get-pip.py"
python get-pip.py
pip install awscli
aws s3 cp s3://%sshPath%/id_rsa ~/.ssh/id_rsa
chmod 400 ~/.ssh/id_rsa
aws s3 cp s3://%sshPath%/id_rsa.pub ~/.ssh/id_rsa.pub
echo "Host *
     StrictHostKeyChecking no" > ~/.ssh/config
git clone https://github.com/zendframework/component-split.git /root/component-split
cd /root/component-split
./bin/split.sh -c %componentName% -r ./zf2-migrate
cd zf2-migrate
git remote add origin [email protected]:zendframework/zend-%componentName%
git push origin master
git push --tags origin
aws s3 cp /var/log/cloud-init-output.log  s3://%bucketBackup%/zend-%componentName%.log
halt
                    

#!/bin/bash
RAM_DISK=3000m
mount -t tmpfs -o size=$RAM_DISK tmpfs /root

add-apt-repository -y ppa:git-core/ppa
apt-get update
apt-get install -y python2.7 curl git php5 php5-curl php5-cli
git config --global user.name "Gianluca Arbezzano"
git config --global user.email [email protected]
curl "https://bootstrap.pypa.io/get-pip.py" -o "get-pip.py"
python get-pip.py
                    

Work in RAM to increase the performance and prepare the environment


pip install awscli
aws s3 cp s3://%sshPath%/id_rsa ~/.ssh/id_rsa
chmod 400 ~/.ssh/id_rsa
aws s3 cp s3://%sshPath%/id_rsa.pub ~/.ssh/id_rsa.pub
                    

Install aws cli tool


echo "Host *
     StrictHostKeyChecking no" > ~/.ssh/config
git clone https://github.com/zendframework/component-split.git \
    /root/component-split

cd /root/component-split
./bin/split.sh -c %componentName% -r ./zf2-migrate
cd zf2-migrate
git remote add \
    origin [email protected]:zendframework/zend-%componentName%
git push origin master
git push --tags origin
                    

Clone the repository with the split procedure, run it and push the new codebase into the component's repository


aws s3 cp /var/log/cloud-init-output.log \
    s3://%bucketBackup%/zend-%componentName%.log
halt
                    

move the output on s3

Warning

@ezimuel is running stats for the new website. #zendframework components have over 52 MILLION installs off of packagist!
@zfdevteam

28€

I know, AWS is expensive!

Thanks

by. @gianarb

https://joind.in/talk/52b11