Ultimate doll 2.0 | cloud native delivery package

I always like some metaphors , In this way, we can understand things more vividly .

Erda It's a PaaS platform , The underlying technologies have been used from marathon + mesos Switch to the current K8s, They are generally considered to be “ Container layer ”.Erda stay “ Container layer ” It's stacked again CI/CD Pipeline、 Cluster and deployment management 、 Application monitoring 、 Automated testing and so on , The embodiment of such layering is very similar to the layering of the network , Each layer performs its own functions , But I prefer to compare it to 「 programing language 」.

Encapsulation is to reduce “ error ”

A high-level programming language has a grammar that is more in line with human understanding , It runs by compiling into assembly or machine code , Or compile it into a low-level language and then compile it again .

A similar ,Erda Through to “ Container layer ” Encapsulation , For our users, it presents “Erda idea ” Function design and use experience . And so called The most important thing about encapsulation is to reduce artificial “ error ”, It's like high-level languages through restricted and elegant syntax 、 Intelligent compilation tips and rich class libraries , Greatly reduce the mental burden of developers , You can easily write robust code . And the programming method above it 、 Best practices , Provide theoretical support for high-speed delivery .

What is a product

Erda Your body bones are made of 「 application 」 For the center , hypothesis Erda If there is only one function left , That is the application “ deliver ”.

Delivery can be very simple , I call it **“ Two steps ”**：

1. Compile the code into an application installation package
2. Install the application in the environment required by the customer

The actual situation will be a little complicated , But it will not change the core steps , It's probably a matter of taking more steps .

If you want to pursue the details , These two steps can be carried out all the way down .Erda The thing to do is to keep users from delving into these details , Just like we don't need to know the encoding when watching video on a personal computer 、 The process of transmission and decoding , There is no need to know the difference between damage and damage .

Of course , Making such a metaphor does not mean Erda Deliberately prevent users from conducting in-depth exploration , On the contrary, if you understand the underlying principles , Better understanding of Erda Understanding of functional design ！

To be specific ,Erda Standardizes what can be delivered on the cloud “ Software installation package ” Format , Such an installation package is called **“Erda products ”**（ Hereinafter referred to as “ products ”）, Let's briefly list the characteristics of the products , In this way, we can have a general impression ：

1. The product is right docker Further abstraction of the image , similar docker-compose.yml, Covers multiple mirrors （ service ） General description of , And the dependencies between them
2. Artifacts are statements of the application delivery environment （ structured 、 Software installation and deployment instructions that can be recognized by the system ）, Not only declare the resource limit of each image service , You can also declare the required middleware （ such as mysql）

It needs to be added , because Erda It is a multi application architecture （ The main library of the core erda、 The front-end application erda-ui、 Monitoring related telegraf、fluentbit etc. ）, therefore Erda When delivering, multiple applications are delivered together , We call it the project artifact . Project artifacts include multiple application artifacts and support batch deployment and other features .

This is a Erda Own products （ Project products ）：

Erda Project artifacts Multiple application artifacts are included by grouping （erda、erda-ui etc. ）

Manufactured Addons details among rds and MySQL They can replace each other

Let's focus on Erda Inside the product of the main application , That is, grouping Group 3 Medium erda Application products of the main library .

because Erda The main application is micro service architecture （ Multi service ）, So we can see a list of microservice containers ：

If the application has multiple services Every service has docker image

And the core dice.yml：

version: "2.0"envs:  ETCDCTL_API: "3"services:  cluster-agent:    cmd: /app/cluster-agent    deployments:      replicas: 1    envs:      DEBUG: "false"    resources:      cpu: ${request_cpu:1}      max_cpu: 1      max_mem: 1024      mem: ${request_mem:1024}addons:  mysql:    plan: mysql:basic

For convenience of reading, we only list the key information , If you need to see the complete dice.yml, Can be found in open source code ： https://github.com/erda-project/erda/blob/master/erda.yml

In this case （ That is to say Erda Own application products ）, You can fully feel Erda How will “ Software delivery ” Encapsulated .

Thanks to the docker Yes “ execution environment ” Encapsulation , And then put it together dice.yml On the service / Multi application integration “ Delivery environment ” Encapsulation （ The number of instances 、 environment variable 、 resource consumption 、 Middleware dependency ）, We can say with confidence ：“ Developers only need to care about what they need to know , The platform is responsible for others ”.

Delivery of products

Although I have just introduced the context of the product , We still want to review the key components of the product ：

1. Image list （docker images）
2. dice.yml Each service is declared （ The image list corresponds to ） Resources required for deployment , And the required middleware

Mirror image （docker） For the time being, let's press ,“ Developers only need to care about what they need to know ” And only dice.yml The syntax specification and configuration of .

understand Erda The implementation of the , You can know that when the platform deploys the artifacts , Read dice.yml Content , And it turns into “ Container layer ” The structure of cognition （ If it is K8s It is Deployment、Service、Ingress, as well as StatefulSet perhaps Operator）, And then turn to “ Container layer ” Deploy . be familiar with K8s Will know , If you let users write these configurations manually , You need to understand a lot of knowledge that you don't know （ Most of them are related to operation and maintenance ）, also It's easy to make mistakes .

PS： But aim at Addons（ Or middleware ） The deployment mechanism of is relatively complex , Consider, for example Rds And other external capabilities provided by cloud manufacturers ,Erda A separate set of deployment and expansion capabilities is provided

As the opening chapter said ,dice.yml It seems to be a “ High level language ”, and K8s yml It is “ Low level language ”（ We are referring to high-order and low-order , Not that “ Higher order ” It must be “ good ” and “ correct ” Of , But for the convenience of human cognition , It is easier to understand and prevent mistakes , And exactly “ The low order ” In performance 、 Flexibility 、 Control and correct logic are more advantageous ）,Erda After a complicated “ compile ”, Will the user （ Let's say business R & D personnel ） Easier to understand configuration and declaration format or syntax , Into actual deployed workloads （Workload） And endogenous or external services （Addons）.

So much talk , It's easy to come to the conclusion that Erda Is the conclusion of how to deliver the product ： One key deployment .

Just select the product , You can create a deployment form with one click , Wait for the result

Of course, the deployment of real production environment is relatively complex , But it will not change the core one click deployment process .Erda In addition, many additional processes and functions have been developed ： For example, artifacts can be exported and imported , We also developed Gallery（ Market ） Facilitate the dissemination of products （Gallery It also carries Addons as well as Pipeline Of Action）; Products are also built migration The ability of , Be able to migrate data during deployment ; wait .

Of course, the most important thing is Pipeline That is, the ability of the assembly line , Through the core of its arrangement CI/CD Logic ：“ Code -> products -> Deploy ”, Be able to control the production environment from the process （ It also includes testing or other environments ） Access to ,Pipeline Of Action The extension mechanism makes it possible to easily connect external process nodes , Of course Erda Built in provides a lot of out of the box capabilities, such as quality scanning 、 unit testing 、 Automated testing, etc .

Last

This article is only from a small side ： products , Tells the story of Erda How to deliver yourself , It also includes how to deliver various other software , but “ products ” Again Erda The most important and core concept in , You could say Erda So far the same “ idea ”. This is a Erda The starting point of the 、 origin , before Erda No body bones , It's just an anonymous internal packaging and deployment platform ; thereafter Erda Numerous boats gather , It doesn't end there .