What is ci/cd| Achieve faster and better software delivery

【 Cloud native 】 What is? CI/CD ? | Smooth delivery obstacles CI/CD

In the previous article, we saw CI/CD How to solve some common problems in software delivery , but CI/CD The birth of is to solve the most essential problem in software delivery 、 The core issue , So what are these hidden problems under the iceberg ？

The essential problem hidden under the iceberg ？

This is actually asking what is the most fundamental goal in software delivery .

The goal of software delivery

An important goal of delivery is speed . The speed of delivery is crucial , One of the important reasons is that we need to verify whether the newly developed functions or fixed defects are really useful . Before software development , Users will guess which functions are needed according to their limited experience 、 Which defects need to be repaired , Therefore, the requirements put forward by users are often not necessarily reasonable and useful . Until users really use the software , These are all untested assumptions . If the delivery time is too long , Then the higher the cost of final correction , This is why we should improve the delivery speed and establish an effective feedback loop .

Another important goal of delivery is to ensure quality . The speed of software delivery is important , But the quality of software is the foundation . The software we deliver should at least meet its business purpose , On this basis, we should deliver high-quality software as far as possible within our capabilities , Bring value to customers .

Combine the above two points , The most fundamental goal in software delivery can be summarized as In an efficient way 、 Fast 、 Deliver high-quality and valuable software in a reliable way .

The goal is ： faster 、 Better

In order to achieve this goal, what should we do ？

Solution ： automation + Frequent delivery

Our solution to this goal is Frequent and automated To release software , This is what CI/CD The core idea of .

First of all, why is automation necessary ？

If you build 、 Deploy 、 The testing and release process is not automated , Then it is not repeatable . Because of the software itself 、 The system configuration 、 Differences in environment and release process , After each of these activities , The results may be different . If each step is manual , Then there is a great chance of making mistakes , And it is impossible to know exactly what has been done . This means that the whole release process cannot be properly controlled to ensure high quality . Automation can avoid these problems perfectly .

Secondly, where is the importance of frequent software delivery ？

If you can publish frequently , Then the difference between each release will be small . This will greatly reduce the risks associated with publishing , And it's easier to rollback , So as to improve the speed and quality of delivery .

Releasing software frequently and automatically is just a macro idea and method to solve problems , What should we do in practice ？ What principles should be followed ？

Principles for frequent and automated delivery of software

1.  Every modification should be fed back

Software can be seen as composed of four parts ： Executable code 、 Configuration information 、 Operating environment and data . Modify any part , May lead to changes in software behavior . So we should be able to control these four parts , Feedback on each modification , Test every change as much as possible .

Four parts that need to be controlled

• Executable code

After modifying the source code , Of course, the executable code will also change . So every time you modify the source code , All need to be built and tested . In order to control this process , Building executable code and testing it should be automated . Each submission builds and tests the application , This is called continuous integration . The built and tested executable code should be used in subsequent deployment activities , Whether deployed to a test environment , Or production environment . Besides , If the application software needs to be compiled , Make sure to use this generated in the build process wherever executable code is needed , Instead of recompiling again to generate a new .

• Configuration information

Any changes to the environment should be managed as configuration information . Under what circumstances , Any changes to the application configuration should be tested . If the user installs the software himself , Any possible configuration items should be tested in various representative environments .

• Running environment

If you need to modify the running environment of the application to be deployed , Then the whole system should be tested in the modified environment . This includes the configuration of the operating system 、 The software set that the application depends on 、 The network configuration , And any modification of infrastructure and external systems .

• data

If the data structure changes , These changes also need to be tested .

Feedback should be done

After detecting the change , We need feedback on changes , Test every change as much as possible , These tests usually include at least the following tests ：

• The process of creating executable code must be effective . This is used to verify that the source code conforms to the syntax .
• The unit test of software must be successful . This checks whether the application behaves as expected .
• Software should meet certain quality standards , Such as test coverage and other technology related measures .
• The functional acceptance test of software must be successful . This can check whether the application meets the business acceptance conditions , Deliver the desired business value .
• Non functional testing of software must be successful . This can check whether the application meets the user's requirements for performance 、 effectiveness 、 Safety and other requirements .
• The software must pass exploratory testing , And demonstrated to customers and some users . This is usually done in a manual test environment . here , The product owner may think that the software function is still missing , We may also find defects that need to be repaired , Also write automated tests for them to avoid regression testing .

2.  Feedback speed should be fast

The speed of feedback is also key . In order to receive feedback quickly , If it is done manually, it has great disadvantages . First, manual operation will take longer , More errors may be introduced , And cannot be audited . in addition , Continue to build manually 、 Testing and deployment are boring and repetitive , Contrary to the guidelines of human resource utilization . We should release our manpower to do more valuable work , Give the repetitive manual work to the machine .

To achieve rapid feedback , The right and reasonable way is to realize automation . The whole automation process includes a comprehensive automation test suite , After many rounds of testing, we have more confidence in the software . If these tests fail , This build will not enter the subsequent stage .

To ensure rapid feedback on changes , In addition, we should also pay attention to the process of developing software , Especially how to use version control system and how to organize code . Developers should frequently submit code to version control systems , Like managing large-scale teams or distributed teams , Divide the code into multiple components . in the majority of cases , You should avoid using branches .

3.  React quickly to feedback

When we capture changes and receive feedback , You should respond to feedback in a timely manner , Adjust actions based on feedback . In this process, we should do ：

Take part in ： Everyone involved in the software delivery process （ Including developers 、 Testers and operators 、 Database administrator 、 Infrastructure experts and managers ） Should participate in this feedback process , This is crucial . If these people can't work together every day , We must often meet and discuss how to improve the software delivery process . For fast delivery of high-quality software , The process based on continuous improvement is very critical . The iterative process helps to establish regularity for such activities , For example, at least one review meeting per iteration , Everyone should participate in the meeting to discuss how to improve the delivery process in the next iteration .

Information broadcast ： Want to be able to adjust actions based on feedback , It is necessary to broadcast the information . Use a large and visible dashboard , Or other notification mechanisms are extremely important to ensure that feedback reaches everyone .

Discipline and planning ： When action is needed , The whole team has the responsibility to stop what they are doing , To decide what action to take next . After finishing this , The team can continue its work .

By automating and delivering software frequently , We have achieved faster and better delivery of software , Gained a lot of benefits , We will be able to verify the changes , Reproduce the deployment process in various environments , To a large extent, reduce the chance of product errors . Because the release process itself is no longer an obstacle , We can deploy software changes , So as to obtain business benefits faster , It also makes software release no longer a stressful process . In the next article in this series , We will learn about CI/CD Bring more unexpected benefits .

Digital China enterprise business group ： What is? CI/CD? | CI/CD Benefits 3 Agree with · 0 Commentary

---

reference ：Jez HumBle,David Farley,《 Continuous delivery —— A systematic approach to distributing reliable software 》

（ Some pictures are from the Internet , If there is any infringement , Delete... Now ）