Imile uses Zadig's multi cloud environment to deploy thousands of times a week to continuously deliver global business across clouds and regions

iMile It is a cross-border e-commerce logistics start-up company , Provide logistics services in emerging markets such as the Middle East and Latin America , Through the self built express team 、 Sophisticated operation and Internet technology , Committed to improving the end of the cross-border logistics .iMile Has grown into the Middle East Top 3 Cross border e-commerce logistics service company , At the same time, it is also the development speed of the local industry Top 2 A logistics company with rich technical benefits .

Pain analysis

In the past business R & D process , We encounter the following pain points ：

1. Environmental governance is complex ：dev、fat、lpt、uat、prod And other data centers in different regions , Use Jenkins Pipeline deployment and delivery requires a lot of manual intervention .

2. R & D efficiency is low ： R & D team program debugging 、 The joint commissioning and testing environment is not friendly , It is often necessary to switch back and forth in different versions of multiple environments to assist in testing 、 Front and rear end troubleshooting , R & D time is occupied .

3. Insufficient test resources ： Scheduled projects and daily iterations are often mixed in the same test environment , It is not efficient to deploy parallelism when a large number of code changes , Affect test progress .

4. Maintenance costs are high ： Service deployment use Jenkinsfile + YAML The way , Each project needs to maintain a set of configurations and scripts , When there are more and more projects , The maintenance cost will be heavier and heavier .

Zadig The journey

Type selection Zadig Before , We have several sets K8s colony , Deployed in different data centers , Use Kubeboard Conduct unified management and control of multiple clusters , There are no other special requirements for centralized cluster management .

encounter Zadig

Inadvertently in B The station saw KodeRover Workathon Share online （ Here's the picture ）, Start working with team members KodeRover（Zadig） This product . It is found that it can be perfectly combined with cloud primitives , Multi cluster services can be managed uniformly , Multi pipeline concurrent construction , Visual delivery of container services , R & D oriented, interactive and friendly .

At that time, our R & D team set up a reconstruction project , A set of independent test environment is required to meet the joint commissioning test task . With a fresh attitude, we deployed a set of Zadig colony , It is planned to solve the test environment requirements of this project reconfiguration . adopt 3 Days of exploration, communication and exchange , We will soon be Zadig The front and rear end engineering services are deployed to K8s colony .

Network transformation

Our network situation was like this ： Development of office intranet and cloud 、 The intranet of the test environment is connected , But to K8s Of Service And the container network layer are not connected .

The refactoring project team started using Zadig The project was tested , But the back-end R & D students found us and put forward a demand “ I think debug This service break point ”. To achieve , You have to K8s The intranet and office network are fully connected , We set out to transform the network , In this way, research and development can be carried out in our own IDEA IDE Call at any time 、debug One of my own services .

A full embrace Zadig

Fully embrace the evolution of cloud native operation and maintenance system , that A full embrace Zadig Is the evolution of the R & D system .

One month later, the reconstruction project went online smoothly , The whole project test is not used with the original test environment NS Isolation , No disruption to the test environment for daily iterations and other requirements , Project deployment becomes very easy , You can view a container log while debugging , The operation experience of R & D students has been qualitatively improved , The original use has been completely abandoned Kubelet The trouble of getting into the container .

With this positive feedback , We immediately made a decision , Will develop the environment 、 The test environment is fully accessed Zadig.

after 4 Months of running in , We will all develop and test the environment K8s The cluster is connected to Zadig. Code engineering passed YAML You can create and deploy the services you need by importing the standard template .

After nearly half a year's efforts , Our research and development 、 test 、 The operation and maintenance team has fully connected all services Zadig. Access Zadig Service near 400 individual , An average of nearly a thousand deployments a week , Greatly improved the R & D efficiency , Let R & D focus on code business realization , The testing team focuses on quality improvement . The test or verification process can pull up a set of environment at any time in a few minutes , For R & D and testing , Reduce the workload and cost of operation and maintenance .

A brief review of several very important requirements details , After these functions are improved , All clusters are connected to Zadig.

1. Because we belong to multi region cross cloud deployment ,Zadig By default, there is only one image warehouse , If we use the same warehouse , The image pulling and pushing of different clusters are carried out through the public network , The pull speed is limited by the bandwidth , And it consumes a lot of traffic .

2. IM Tool message prompt push text optimization .

3. Granular control of project authority management .

Overall revenue

Zadig adopt “ workflow ” The whole is connected in series K8s Components of , It also connects our entire R & D team , It greatly reduces the maintenance of scripts 、 Environmental governance , At the same time, it is also very simple and efficient . It is of great help in project iteration and delivery , It saves a lot of time and cost , Let professional people do “ major ” Affair , Make the project R & D efficient and concurrent , Reduce communication between teams Gap, It is very helpful for our R & D delivery .

The conclusion is Simple , Efficient ！

5、 ... and 、 Expectations and suggestions

1. Service image version rollback , At present, there are only local clusters （Zadig Deployed clusters ） You can use the mirror version to rollback , adopt Agent The connected cluster cannot rollback the image .

2. Refine the granularity of permission control , You can customize permission groups or services to users or user groups .

3. Support multiple deployment methods , for example Android Native APP The construction of the project , We try to build by customizing the image , But Android relies heavily on resources , The image is also very large , It takes longer to pull the image and start the image than to build it directly on the virtual machine .

4. Expect to test functions and API A richer set of functions , You can consider improving the plug-in mode Zadig The ecology of .

thank Zadig The team's constant help , During this period, we discussed many requirements together , Have been quickly responded to and resolved . Look forward to Zadig You can go abroad , To the overseas market ！

Zadig, Let engineers focus more on creating . Welcome to join Make complaints about open source Tucao group

Zadig on Github
Zadig on Gitee