[beauty of software engineering - column notes] 26 | continuous delivery: how to release new versions to the production environment at any time?

---

Column information

1. Columnist ： The treasure jade

2. Column publishing platform ： Geek time

---

primary coverage

As of today , Continuous delivery has become a recognized good development practice , More and more development teams have applied continuous delivery , It is automated , Let your code after each submission , Can automatically complete the compilation 、 Testing process , After completion, you can be ready to deploy and release .

Continuous delivery if subdivided , In fact, it can be divided into Continuous integration 、 Continuous delivery and continuous deployment Three concepts .

---

One 、 Integrate 、 History of deployment and delivery

1. The evolution of integration

   When developing multi person software projects , Everyone will be responsible for the development of some functional modules , Integration refers to the branch code developed by everyone , Merge into the trunk , In order to test packaging .

   • The original stage of integration

     Back in the age of waterfall development , In the development stage , Generally, it is not integrated . Everyone develops independently , Wait until the development phase is almost over , Then submit the code to the source code management tool , Let the code integrate , To compile 、 Deploy and publish to the test environment .

     
     《 restructure 》 The author of Martin Fowler Said ：“ If one thing is painful , Then do it more often （if it hurts, do it more often. ）”, Continuous integration is also a painful thing to integrate , Do it more often .

     

     The benefits of continuous integration are obvious ：

     1. Cooperate with automated testing , This ensures that the trunk code is stable .
     2. Frequent integration allows developers to always get the latest code from the trunk in time , Not like a class library 、API Inconsistency and other problems were not exposed until the final test stage .

2. History of deployment and delivery

   Deployment refers to publishing code to various environments , For example, deploy the test environment for testing . Delivery refers to the software product after passing the test and acceptance , Have the conditions for release to the production environment and delivery to customers .

   1. The initial phase of deployment and delivery

      In the early years , Deployment is a troublesome thing . You need to get the latest source code manually 、 compile 、 Many configurations need to be modified for the environment .

   2. From manual deployment to script automated deployment

      For programmers , If one thing can be solved automatically , Sooner or later someone will find an automated solution , Deployment has also evolved from manual deployment to Automated Deployment .

      Early Automated Deployment solutions were Daily construction （Daily Build）, Simply speaking , That is, after work every night , The daily builder automatically downloads the latest code from the source code manager , compile 、 Deploy the program to the test environment .

   3. From script deployment to continuous delivery

      In fact, after understanding continuous integration , It's much easier to understand continuous delivery . Continuous delivery , On the basis of continuous integration , Further , After the functions are merged into the trunk , Not only automated testing , Also packed , And deploy to the test environment .

      Theoretically, it can also be deployed directly to the production environment , But this link needs to be confirmed manually .

      

   4. From continuous delivery to continuous deployment

      Continuous delivery , For the deployment of production environment , Manual confirmation is still needed . And continue to deploy , The only difference from continuous delivery , There is no manual confirmation , Every time the code merges from the branch to the trunk , After the automation test passes , The production environment will be deployed directly and automatically , No manual confirmation is required .

      however , Continuous deployment should be done well , It's still a challenging thing , After all, from code consolidation to production environment deployment , If there is no human intervention, only rely on automated testing , This requires high coverage and stability of automated testing .

---

Two 、 Should the application be delivered continuously ？

We are waterfall model development , Should the application be delivered continuously ？

The answer is ： Continuous delivery doesn't matter what development model you use , Waterfall model can also be applied to continuous integration , The continuous integration environment and corresponding development process should be set up as soon as possible , You can see the benefits immediately .

1. Expose problems as soon as possible ：Martin Fowler Said ,“ Continuous delivery cannot eliminate Bug, It's about making them easy to find and correct .”
2. Greatly improve efficiency ： Continuous delivery allows the development process to merge from code , Until the final deployment , Have achieved Automation , Can greatly improve efficiency .
3. Improve quality ： You need to pass automated tests before each merge , So there will be fewer mistakes .
4. Reduce project costs ： When we first set up a continuous delivery environment , It is necessary to invest a certain cost , But in the long run , Development efficiency has been improved , Code quality has improved , Instead, it is helpful to reduce the overall cost of the project .

---

3、 ... and 、 How to build a continuous delivery environment ？

1. preparation

   To build your own continuous delivery environment , It's not easy to find a continuous integration tool to work , But we still need to do some preparatory work .

   Let's first look at the continuous integration part , Continuous integration is relatively simple ：

   1. Need source code management tools , for instance git、svn, Because continuous integration tools need to get code from a unified source code warehouse ;
   2. Need to write automated test code , Because there is a very important condition for continuous integration , The automatic test must pass ;

   Continuous delivery is relatively more demanding than continuous integration , Your project needs to meet the following conditions ：

   1. The process of building code can be repeated , And the results of each build are consistent 、 The stability of the ;
   2. The configuration of all environments exists in the source code management tool , It's not just code ;
   3. You need to automatically create distribution packages for different environments ;
   4. The deployment and release steps of all environments must be automated ;

2. Choose the right continuous integration tool

   There are already many options for continuous integration tools , There are open source 、 commercial , Hosted on cable , And I built it myself .

   The main continuous integration tools are these ：

   • Jenkins
   • Go CD
   • Travis CI
   • GitLab CI
   • Azure Pipelines

3. Implement according to the selected tool

   After choosing the continuous integration tool you want to use , You need to build according to the instructions of the tool . This part is relatively simple , There are also many tutorials online .

---

Four 、 Other excerpts

• Before the code is integrated into the trunk , Must pass automated tests , As long as one test case fails , You can't integrate .
• Although continuous delivery is not yet popular , But the future is like source code management , Become the standard configuration of the development team .
• The main problem of continuous integration is to build the whole continuous integration environment , A little more trouble , In addition, it needs to cooperate with some process specifications to assist the implementation , For example, a certain coverage of automated test code is required , Only after passing the test can it be merged into the trunk .

---

summary

Build a continuous delivery environment , First of all, we need to be prepared , For example, automated test code and Automated Deployment scripts ; Then choose the continuous integration tool ; Finally, implement according to the selected continuous integration tool .