Software is great and all. But truly nothing matters except delivery. This course is intended to teach you the philosophy of Continuous Deployment. Continuous Deployment refers to automatically making any submitted code available for testing every time a code change is submitted. Many engineers today believe that this method results in more reliable code. In addition to teaching you about Continuous Deployment (or Continuous Integration), this course will also teach you to create a deployment of one of your Lambda School projects using this method.
DevOps is the complicated framework that surrounds transmitting your code from your local machine to the public. The specific steps taken to perform this transmission (aka: deployment) are extremely varied. There are many dozens of web sites dedicated to helping you deploy your code to the internet and collect a small fee from you for this service. Regardless of which service or technology you use, DevOps is always loosely comprised of the following steps:
- Connecting a computer to a network (usually the internet).
- Installing and configuring an operating system on that computer.
- Installing software on that computer that can open a socket and respond to incoming traffic.
- Customizing the above software to perform a specific business, entertainment, or hobby-related task.
- Configuring the above software to become failure tolerant (automatically restart).
- Configuring the above software to be data-consistent.
The above steps are sufficient to take your local software development project to the public. You can trivially hire a service to perform the first 3 steps, leaving you only with the remaining 3. You can ignore steps 5 and 6, leaving them up to the software configured in 2, and you can create static websites that aren't subject to the issues in steps 5 and 6. Geocities, Wordpress, Small Victories, and many other sites allow you to implement only step #4. If you wish to be a professional web developer, you will need to learn the rest of the steps.
Finally, the above steps only create a relatively brittle site. In particular, these steps don't help at all with the development process - that is, the process of writing and refining your application code to become reliable, feature-rich, and bug free. These steps in particular do not scale well for projects that are not toys: most development projects involve many developers, technologies being developed in parallel, and thousands or millions of expectant customers waiting for each revision. In order to safely and effectively deliver code to the world at large, more development is needed.
The above notes about DevOps.
- Domain names - ICANN and Namecheap, WHOIS, A records
- Security - HTTPS and Digicert, self signed HTTPS
- Scalability - Dedicated hosts versus your own hosts
- Name servers and virtual hosts via CPanel and SSH (Godaddy, Geocities)
- Super lightweight hosts like Small victories
Now, get ngrok and start using it with the free account. Choose a personal or Lambda web site project that you'd like to share with the rest of class (and the world), run it on your local machine, and share it with us using ngrok.
The rest of this document
Any sufficiently advanced Ops is indistinguishable from Dev
- Baron Schwartz @xaprb -
- Steps 1 and 2 from above can be automated to a degree
- Write software that loads software from elsewhere
- Test the software as soon as it is loaded from elsewhere (unit and integration tests)
- Build/compile/deploy the software to the same computer or another computer
- Report on test and build results to another computer automatically
- Monitor the stability of the deployed software
- Add more software when traffic increases, remove it when it decreases (Load balancing)
- Hosting servers in different physical locations in order to improve performance and fault tolerance
- Splitting up data in different locations to improve performance and fault tolerance
Topics:
- Hosting and network/device security
- Ngrok
- Static hosting
- Shared environment hosting
- Dedicated server hosting
- Front end hosting
- Database hosting
- Cloud hosting
- Microservers
- Different servers for different jobs
- Front end server
- Database server
- Authentication server
- Micro services
- Integration server
- Web servers(nginx) and simple servers(express)
- Google, AWS, Heroku, Azure, etc
- Containerization
- Automated test
- Automated deployment Whenever you push to a specific Git branch, your integration server receives a Git Hook, requesting that the repository be built, tested, and deployed.
- Versioning Every piece of software is running a specific version. Version numbers actually reference a specific container configuration and a specific git commit version.
- Manual Testing
- Security in motion (HTTPS)
- Security at rest (ssh, pub/priv key encryption, MFA, password danger, etc)
The following assignment is extremely difficult. Do not lose heart if you are lost or struggling. In professional work I would allow between 1-3 weeks of 8 hour days of development to accomplish the following goals.
Let's build a CI pipeline:
You need:
- A front end in a repository
- Unit tests for your front end
- A back end in a repository
- Integration tests for your back end
You will: Follow instructions at GCP for creating a Docker container for your repository
Google Cloud Platform to host a back end: Quick GCP setup NOTE: When going through this GCP tutorial, it will ask you to enable billing, meaning you'll need to give Google your credit card info. The tutorial also says that this tutorial will cost you a few dollars. However, new GCP users are eligible for $300 in credit, which you should take advantage of.
Your back end will be an express server with two endpoints:
/version - returns the current software version
/time - returns the current time
Ignore everything about the slack bot, except that kittenbot.js will actually be an express server. Do not try to create a slack bot or get its access token. We're using this tutorial because it builds a GCP container very quickly.
Then, test your docker container with the command line:
- Is it running? (curl localhost:port)
- Can you stop it (docker command)
- Can you start it again? (docker command)
Google Cloud platform to host a front end: Quick GCP setup
This is the same tutorial as for the back end. Ignore the stuff about the slack bot. Instead, you will check out this repository to the docker container.
Set up Jenkins on Container Engine
Your front end will be a React app that displays its own version, the back end's version, and the time from the back end.
// Reference commands, you will find them in the above tutorials.
brew install docker
do (https://cloud.google.com/sdk/docs/#mac)
./google-cloud-sdk/install.sh
gcloud init
gcloud config list
gcloud components install kubectl
create project at (https://console.cloud.google.com)
after created, check Container Engine, Container clusters, Container Registry
gcloud config set project lambda-XXXXXX
gcloud container clusters create lambda-cluster
gcloud config set container/cluster lambda-cluster
gcloud container clusters get-credentials lambda-cluster --zone us-central1-a --project lambda-XXXXXX
kubectl proxy
http://localhost:8001/ui