Cloud Native Computing Foundation KubeCon + CloudNativeCon China 2019 (Shanghai), 26 Jun 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Lightning Talk: Developer's Toolbox for Testing in Production - Rui Chen, Meetup

Description

Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io

Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects

Lightning Talk: Developer's Toolbox for Testing in Production - Rui Chen, Meetup

In the software development lifycle, test pyramid helps to build confidence for launching the feature. While single code change is easy to manage, testing a whole feature in the system are facing the increasing challenges in the microservice world. While maintaining multiple environments would help sort out the promotion path, but it is involved with tremendous efforts cross the teams and very costly for the business as well, that is how "Testing in Production (TIP)" becomes popular. In this session, I want to talk about some of the TIP practices that we do at Meetup for delivering features. Especially think about the feature delivery from developer experience perspective.

https://sched.co/Nrs6

A

So today, I'm going to talk about the topics about testing production. This is the emergent op when we actually bring the DevOps into the DevOps culture into a company and I actually had a subtitle according moving the needles. What drive the technical change in inside a company? We done the user, no like a vendor law. Like other infrastructure engineer, I would like to think a vacation development. In a different perspective of the modern avocation stack, we move, firstly, move my mollusk application to cloud. We actually leverage different cloud vendors to build native web apps.

A

We micro-service given backends, we build it. Sp applications on s3 were different kind of patches and was a leveraged CDN to talk to the what else up across the world. So the applications target, search and wisdom mindset. We needs to rethink the fail, faster philosophy which is core to the agile process. This is very. This is the news. This is a new concept, the way to rethink the same existing in a different world, so that website really has two camps. One years they've had the otherwise option with them. Have you actually were thinking? Go darker?

A

Go kubernetes coast of smash, but with warps head, you always think about the CVG problems or you think about the capacitive planning, where you think about the unco issues every single day, and we also like we talked about their city, which I did mention last year. The clcd is just like putting what is testing efforts to shipping products in a faster ways right, so you have unit as in the in the in the in around, and you have component test integration has a theaters in the middle I have good house of.

A

Sometimes you also have many testing to find out the box. When we talk about testing production. This is critical because every single piece is going to sitting in one environment, but we also have developed tools to facilitate of the testing process as well, because we have logging, we have monitoring. We have alerting with loes tools this kind of what a girl can be shipping into one parameter.

A

So testing production is about just one environment, about every services, where every every every everything, because you don't need to have self added tools to talking to your different environments, you don't need to have one dashboard to include so many different pieces. When you have problems when you have uncoil issues, you don't know what is the best of metrics to communicate out and also with this one environment, you can involve your tests to better suit for your kids, for your application.

A

Development think about the test is what so-called code needs maintenance that actually brings the feature flags. What is the feature feature? Flexi spot? You actually have the things that you want to roll and the same that you won't don't want the diffusers to see it's about. Versioning is about routing or logic, to write back end to the right source of truth, so features like one one years just to have the raw to defend account. For example, in our case, we have feature text calculations to some some of the countries we need to you.

A

Leverage IP address to find out the country as to determine official is silve all the user or not. Sometimes, you have gdpr issue that you don't want to go all the features, kill a European audience and T's, and you until you figure out the PR problems, the the other most uses. Your feature. Flexi is about cleaning features you of legacy code. You have moralistic applications called sitting in the backend for like 10 years. He wants to kill them, but nobody in the team under standard code.

A

So you want to put a feature flag, wrapped up to make sure that code can be saved to kill, and there are other way to think about the visual flair keys. You can wrap the features depending on your use case. Sometimes the features around members sometimes in the feature, is around a high level and member. You can also use feature flags to do that.

A

The canary release is about traffic migration so that we can actually have pipeline to shift to the traffic group from production to the shadow traffic feature flag is about tangling wolf canary is about the pipeline work to have the pests which are tested against the different themes a pipeline.

A

The final piece I wants to talk about is the logs metrics and choices you're, probably going to hear a lot in other edge situations, but accountability is logs about aggregation time series view the metrics is about performance is about testing the certain keys of action code. Assistance is the easy. Easy way is keen point down the different, the transaction of macro service. In large picture we wasn't leverage to open to will have also as leverage pieces between front and the back end, so that we can identify the issues for the users.

A

So when we actually test in production, you need to consolidate understanding. You need to evolving the dashboard for the better use case to find out opponent lack and the we stab with that we can actually involving the the the practice in the in the long run. That's what I want to share today. Thank you.

A