Posts

Showing posts with the label DevOps

[读书笔记] 是时候埋葬DevOps了

https://mp.weixin.qq.com/s/CoULPbCd8xZVEU19t7oliw   建立“DevOps”团队的公司正朝着正确的方向前进,但它们需要从基础架构配置管理转向平台工程和使开发人员能够实现自助服务。 知识孤岛和专业知识是同一枚硬币的两面。从全栈工程师到DevOps从业人员,我们行业喜欢假装每个人都能做任何事情。我们这个行业的人生性喜欢捣鼓。我不知道我们是在自欺欺人,还是这个行业一直在利用我们喜欢捣鼓的天性,但是时候将DevOps扔到窗外了。 越来越流行的时代精神是“平台工程是未来。” 遗憾的是,许多组织无法指望“DevOps”团队来搞平台工程,从而抵达成功彼岸。

读书笔记 - 2023 State of DevOps Report Platform Engineering Edition

Image
A product manager role on the platform team is crucial to cultivating a product mindset , evangelizing the platform , and effectively disseminating information throughout the firm.   ---  including i ncreasing awareness of platform capabilities (47%), setting realistic expectation s for the platform team’s role across the organization (44%), and more c losely following industry trends and keeping up with feature requirements (37%). What is Platform Engineering? Platform engineering is the discipline of designing and building self-service capabilities to minimize cognitive load for developers and to enable fast flow software delivery. Platform teams deliver shared infrastructure platforms to internal users responsible for delivering a value stream – typically software developers and engineers. The platform team continuously develops, builds, maintains, and supports underlying infrastructure in order to build self-service solutions, enabling development teams to deli...

DevOps 缺少定义,平台工程需要指导性路线图

https://www.infoq.com/articles/platform-engineering-roadmap/ Key Takeaways A lack of definition for DevOps enabled early adopters but didn't allow late-majority enterprises to be successful in their adoption of DevOps. The platform engineering community is in danger of repeating this mistake What is missing is a map for how to progressively adopt a Platform Engineering approach, not a highly specific end goal While Team Topologies has provided an excellent starting point, not everyone understands it is more than just implementing the right types of teams A common anti-pattern is setting an open-ended goal of increasing collaboration between teams. This tends to be highly inefficient at scale as stream-aligned teams outnumber platform teams. Most enterprises fail at transformations like this because they implement them via project management instead of product management At PlatformCon 2022, Nigel Kersten, Field CTO at Puppet, spoke about the need for a prescriptive roadmap for plat...

[读书笔记] 管理宠物到管理牛群,DevOps 场景下效率难题如何解决

Image
https://www.infoq.cn/article/jTJGTtu2AgX74GkGif8Y 整个 CI/CD 过程就是一个不断将可变变为不可变的过程。 每一步都是要锁定一部分内容,做到不可变。有了不可变,才能随时得到不同资源组合的稳定版本,然后才能重复部署,或者安全及准确的回退到之前的版本。 组织团队之间的通讯和交流的模式,最终会映射成软件系统或者产品的形态。 独立部署是微服务的精髓,服务部署时还要依赖其他团队,本质上就不是微服务了 ( 不过是一个很昂贵的解决方案 ) 两个 container 可以共享一个存储,叫做 volume;通过 Init container 把数据准备到这个共享存储空间的路径上,Main container 就可以在这个存储上加载数据。这样一来,Main container 就会与数据加载以及初始化的方式彻底解耦。 如果在 pod 中包含多个 container,当我们需要某个 container 先完成启动就绪,就绪完成后才继续下一个 container 的创建,那么就可以在前一个 container 中加一个 postStart。这个 postStart 过程只是检查自己是否就绪,如果一直没有就绪,退出后返回不正常时整个 Pod 会失败,无法创建成功。如果创建成功,意味着已经就绪,就可以继续启动下一个 container。 当一个 Node 上有很多 Pod 在运行,可能会出现 Node 硬件不够用的情况,那么哪些 Pod 会被杀死?哪些 Pod 会被保留下来?这个叫 QoS。 Affinity 亲和性和 Taint 这两个概念有很多人都搞不清楚。所谓亲和性是从 Pod 的视角去看,什么样的 Node 适合我来运行。Taint 是从 Node 的角度来看自己可以运行什么样的 Pod,兼容了 node 上的 taint 的 pod 才可以被调度到 node 上。 引入了容器之后,带来的改变绝对不仅仅是构建、部署和维护的方式这么简单。其实每一次软件设计的变更都涉及到新增的构建块,带来新的设计思考。

如何利用开源框架实现运维编排 - 读书笔记

Image
 

DevOps平台的产品规划思考 - 读书笔记

Image
 https://www.toutiao.com/a7043600901508088334/?log_from=94eb1c45d0dc5_1640052448731&wid=1640288269363

Operationalization in Data era

A few challenge on AI/ML/DS How the data scients to cooperate with developer, what are the role of each other? How to comply with enterprise security policy What is the Devops for DS,  ML is differnt from tradnitonal IT,  don't let business to deploy on the platfrom Operationalization Pipleline the DS team What is the platform for it Scalable,  the code runs on laptop should be able to be parallelly run on cluster.   Should be done on code level, use for example Dask,  and platform level,  use Kubernates.  

DevOps Tools ,and its test process

Image

An example of Continous Development

Image
https://medium.com/@ahmetatalay/continous-deployment-via-gitlab-jenkins-docker-and-slack-5d08836d01e0#.r1b33bco2

Cost of Test

Image
Test is expensive, the TA's job is to achieve higher quality with lower cost. It is difficult, but with subtle design, we can somehow achieve it. It is not risk-free, but try to manage the risk, and cost.  Control the false alarm which is a big cost in the test project. 

AutoTest Flow in DevOps world.

Image
From post:   https://goo.gl/YXVctP

QA in DEV-OPS

A few good point:  1. Dev-DevOps views vs QA DevOps view Then How to cut the DevOps Views:   Move everything to happen earlier. Everything is code, Everyone should be technical.  test constantly. Relentlessly Automate is a key to achievement Team to have a  broaden knowledge.   Follow the Discipline and Create Visibility.  All in All we would like to get Higher Productively .