Feature toggle

Introduction: Currently, almost software is running stably, but users want to upgrade some functions or develop new features. After deciding to do a new function, the installation and testing process took a long time to ensure that the new function works well and matches the needs of the user, without having to change too much.

Introduction:

  • Currently, almost software is running stably, but users want to upgrade some functions or develop new features. After deciding to do a new function, the installation and testing process took a long time to ensure that the new function works well and matches the needs of the user, without having to change too much. After developing the new feature, we will let some users use this function, others will still use the current version of the application.
  • In the process of developing new features, still ensuring the development and upgrading of old functions


A simple way to do this is create a product branch that goes from the current branch and develops in parallel into 2 different products, but then managing and merging the code will be complicated and also very difficult. to manage as well as cost to develop 2 versions at the same time.
Thus we arise a requirement that allows the developer to change the operation of the application, the flow, the interface or the processing features without having to modify the code. To make it easy to imagine it as if you use facebook with many features that it is only suitable for a certain number of users, or software versions with advanced features for paid users…
The problem sounds complicated like how to determine which users are used, which functions are used or not, how data is stored and returned to ensure security and not affect the data. Current function…
However, we can divide it into many levels to implement this function such as installation at the application interface, installation at the api layer, installation at the database layer and deeper at the hardware level.

So to store this information, where will we store and it, what solution:

  • Solution 1: The flags are stored in the code, each change needs to rebuild the source code and deploy to the end user, this way will cost money and effort to deploy
  • Solution 2: The flags are stored in the application’s config, for example, config on a web server, web hosting, app config, and env in node js code … this way will require the user to reload the application and will be difficult to manage
  • Solution 3: Self builds a feature flag system, build a flag admin portal, and the necessary services to operate a flag management page
  • Solution 4: Use the feature flag service from an external service.

Solution

Solution 1

For basic functions or do not need to be turned on and off often, not applicable to different sets of customers, so the setting here is quite simple. You can create a flag in the source code to turn it on and off.

functionshowWelcome(){// ... old UI welcome}functionshowWelcomeV2(){// Show UI welcome version 2, work in progress}// Old code before setup feature flagfunctionshowUIWelcome(){showWelcome()// Show UI cũ}// Feature flagfunctionshowUIWelcome(){// config feature flag SHOW_NEW_WELCOMEconstSHOW_NEW_WELCOME=true;if(SHOW_NEW_WELCOME){showWelcomeV2()}else{showWelcome()}}

Solution 2

For functions that are likely to change periodically or under certain conditions that you can manage or plan, you can save those settings in an environment variable. In today’s source code platforms all allow you to configure environment variables dynamically and allow to relaunch or reload environment variables dynamically. In addition, this approach is suitable for small systems, the operator has good system administration ability. However, it is important to pay attention to security requirements when setting flags as well as re-testing

  • Advantages
    • Simple, easy to deploy with most platforms
    • Fast deployment time
    • Low deployment cost
  • Defect
    • Limited features, It only responds to turning on and off functions in a basic way
    • Does not guarantee security requirements

Implement guideline:

Create config in env something like that:

SHOW_REFERRAL_SCREEN=true
SHOW_MENU_SCREEN=true
SHOW_RESET_SCREEN=false

function showUIRefferral() {
    if(env.SHOW_REFERRAL_SCREEN){
        // ... UI referral
    } else {
        // show UI empty
    }
}

Solution 3

For medium to large projects with a good development team wishing to build feature flag functionality for long-term use with our products and services, the self-development option should be implemented. To develop a module like this the team will need to very clearly define the needs of the service, and analyze the fully functional design.

  • Advantages
    • Actively develop according to demand
    • Easy to upgrade according to the design of the project
    • Our system can scale and maintain easier
    • Cost won’t increase so fast after our app has been scale
    • Developers can merge and release their new feature code without waiting too long to be merged, and user experiments won’t be affected after release ⇒ CI/CD
  • Defect
    • Development costs
    • Functions are not rich
    • Long development time
    • Cost of personnel to operate the infrastructure system, automatically scaling

Implement guideline:

Approaches

  • Implement authorization system with Module based for none authentication application
  • Implement authorization system with Module based plus user role based if needed for authentication application

Pros

  • We can easy to custom with more complex requirements later
  • Easy to scale and decrease cost when compared to 3rd party

Cons

  • Take time to init and implement in DevOps, Backend, Frontend, and Admin to manage
  • With small we can’t hit deadline or we have to accept tech debt later

What could go wrong?

  • We can take too much time to implement our own system and get results not as expected when lack resource and time

Design own system

Database

Database name: Authorization
List of tables

Diagram for feature flag
Diagram for feature flag

System architecture

System diagramInfrastructure

Backend

Database migration

Follow the diagram above to write a migration
API endpoint

  • CRUD for Admin
    • Need to broadcast to socket channel on each action
  • get-enabled-feature-flags endpoint to return list to FE App based on authentication

Socket

  • Enable socket to help FE join to channel and listen topic
  • Send socket event to FE after changing anything effect to feature flag

Middleware

Input for middleware is a list of conditions with condition key and value, we will implement to get list of feature flags are enabled for these conditions

Flow to get feature flags base on condition inputs

Flow to check condition is true or false in detail

Example

Old men, children, and women OR man between 20 → 40 can see feature flag A

  1. Create feature flag: Name + Key

  2. List and select existing triggers if not

  3. Create the first trigger

    1. List variables

      Name Variable Key Variable Type
      Age age number
      Sex sex enums
    2. List condition

    Name Variable Key Condition Type Value
    Children age less_than_to_equal 16
    Old age greater_than_to_equal 60
    Woman sex equal woman

    c. Create trigger: Name + Description + list conditions

  4. Create the second trigger

    1. List variables

      Name Variable Key Variable Type
      Age age number
      Sex sex enums
    2. List condition

    Name Variable Key Condition Type Value
    Man Min age less_than_to_equal 40
    Man Max age greater_than_to_equal 20
    Woman sex equal man

    c. Create trigger: Name + Description + list conditions

  5. Add these triggers to the feature flag A

Middleware

  1. Get information
  2. Create list input conditions want to get feature flags: [{key: “age”, value: 35}, {key: “sex”, value: “man}]
  3. SELECT * FROM variables v WHERE v.key in [”age”, “sex”]
  4. SELECT * from conditions c WHERE c.variable_id in [variable_ids]
  5. Get a list of true conditions
  6. SELECT * from triggers_conditions tc WHERE tc.condition_id in [true_condition_ids]
  7. Loop triggers → filter trigger has all conditions in the trigger are a subset in true_condition_ids
  8. SELECT * from triggers_flags tf WHERE tt.trigger_id in [enabled_trigger_ids]
  9. Return list of enabled unique feature flags

Admin system

For Homepage

For create feature flag

For create trigger

Frontend Apps

Prerequisites

  • The backend and Frontend need to define list of feature flag key before. If we define before we can implement independency

Implementation

  • FE Call to BE to get the list of available feature flags
  • Save a list of available feature flags to context state or local storage
  • Create util function isAllowFeature(’feature_key’) → rendering condition

Solution 4

For small and medium projects or MVP products that are in the process of testing, it is recommended to use support solutions from appropriate service providers to optimize development time and operating costs. as well as can quickly use many outstanding features compared to installing it yourself

  • Advantages
    • Low service cost if there are few people using the service
    • Many advanced functions
    • Rapid deployment
    • Multi-platform support with SDK
    • Real-time toggle feature
    • Support decentralization of management
  • Defect
    • Service costs can be high when scale many users using the service
    • Unable to design the specific requirements of the project
    • Can’t solve some special requirement when improving feature

Implementation guidelines:

Setup code on frontend for each feature

// To use in your project, import the initialize functionimport{ initialize }from'@devcycle/devcycle-js-sdk'// The user object needs either a user_id, or isAnonymous set to trueconst user ={ user_id:'my_user'}let dvcClient

try{// Call initialize with the client key and a user object// await on the features to be loaded from our serversconst dvcClient =awaitinitialize('dvc_client_f895833e_2f80_4631_828f_e2fd5dc0f27b_23918a1', user).onClientInitialized()useDVCVariable()}catch(ex){
  console.log('Error initializing DVC: ${ex}')}functionuseDVCVariable(){if(!dvcClient)return// Fetch variable values using the identifier key coupled with a default value// The default value can be of type string, boolean, number, or objectconst dvcVariable = dvcClient.variable('liquidity-pool',false)if(dvcVariable.value){// Put feature code here, or launch feature from here}else{// Put feature code here, or launch feature from here}// Register a callback to be notified when a value is updated
  dvcVariable.onUpdate((value)=>{// updated variable value is available})}
  • Create a new feature flag, Select

  • Define feature flag name

  • Add a condition to feature flag

Solution 1: Config flag in source code Solution2: Setup config in the environment Solution 3:Build a feature flag management system Solution 4: 3party Feature flag service
When implementing this solution – The whole team wants to enable the new feature flag and let developer turn off it if needed – When the team can change env on the application easily so we can switch feature flags without rebuilding and touching to code – One source code and different feature flags status with multiple servers – Want to dynamic feature flags with multi variables – Feature flags rules and status changes regularly – The team has enough time and resources to implement it – Want to dynamic feature flags with multi variables – Feature flags rules and status changes regularly – Team willing to pay to save development time
Don’t require DevOps (difficult to deal with a big team) ☑️ ☑️ ☑️
Other teams can switch feature flags ☑️ ☑️
No cost when scaling ☑️ ☑️ ☑️
Save implement time and resource ☑️ ☑️ ☑️
Custom dynamic and complex feature flag conditions ☑️ ☑️
Use case – Show banner or new form until marketing team wants to turn off it – Enable feature flag base on environments like sending login email on production but turning off on development and staging – Want to dynamic feature flags with multi variables

Nguồn: viblo.asia

Bài viết liên quan

WebP là gì? Hướng dẫn cách để chuyển hình ảnh jpg, png qua webp

WebP là gì? WebP là một định dạng ảnh hiện đại, được phát triển bởi Google

Điểm khác biệt giữa IPv4 và IPv6 là gì?

IPv4 và IPv6 là hai phiên bản của hệ thống địa chỉ Giao thức Internet (IP). IP l

Check nameservers của tên miền xem website trỏ đúng chưa

Tìm hiểu cách check nameservers của tên miền để xác định tên miền đó đang dùn

Mình đang dùng Google Domains để check tên miền hàng ngày

Từ khi thông báo dịch vụ Google Domains bỏ mác Beta, mình mới để ý và bắt đầ