Ceph Cephalocon Barcelona 2019, 24 May 2019

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Using Devops Practices for Operating CEPH - Anders Bruvik, Safespring

Description

Using Devops Practices for Operating CEPH - Anders Bruvik, Safespring

Devops is a set of practices that aim to bring together Developers and IT Operations, and the goal is to shorten development lifecycle and bring IT in closer alignment with business goals.

So does it make sense to talk about Devops practitions when the "only" thing we do is operate a storage cluster like ceph? Yes - because modern system operation is becoming more and more about development - infrastructure as code is a term that describes how we increasingly are using development practises to configure IT systems.

In this talk, I will discuss why and how Devops matter for IT operation teams, and I will illustrate by bringing in examples from our experiences in building a distributed hybrid storage cloud in Sweden.

About Anders Bruvik
Safespring
Infrastructure Engineer
Oslo Area, Norway
Websitehttps://bruvik.me
Twitter Tweet LinkedIn Connect
Anders is an infrastructure engineer at the Nordic infrastructure provider Safespring. Before that, he spent years in different technical and management positions at a large Norwegian university where he worked with everything from config management on Unix servers to end-user computing, mobile platforms, project management to virtual desktop infrastructure. When he’s not engineering stuff, he spends time organising DevOpsDays in Oslo and Stockholm.

A

Okay, you don't need devs to do the ropes so I'm on a three-week I work as an infrastructure, engineer and I'm, not a developer, as you might have guessed, I. Also organize a couple of the upstairs so I'm going to talk about how we can apply DevOps practices when you are working in an infrastructure team, but first I'm going to try to explain what DevOps is, and it's not that easy to explain, because everyone tends to have their own definition of DevOps. So I'll start with telling you what it's not so. Is there anyone here?

A

Who has the title DevOps engineer as the job title? Sorry, you've been fooled: it's is there anyone who have a DevOps team in their organization? Sorry, it doesn't exist.

A

Actually, there's a lot of things that the office is not it's not a product, it's not something you can hire. It's not a department. I asked a question out on Twitter and I got a lot of different and some funny answers. It's not something that anyone understands so my definition of DevOps, it's something you do I want to share this.

A

This definition from ken McGraw DevOps. It's a culture where people, regardless of title of background, work together to imagine, develop, deploy an operating system and and when you start to talk about DevOps as verbs instead of instead of a thing, it starts to make more sense to talk about the hoops so and we are in an infrastructure team because after all, we are developing our infrastructure or search clusters. So if it becomes developing and operating it makes more sense. So and this definition it's from John Willis and there's a great blog post.

A

You can read so.

A

What is this thing we do when we do DevOps I would say the most important practice is to make pipelines and what is pipelines it's? What takes our configuration from version control into production, so we have worse in control and we have something that automatically brings it into production, and why do you want to do that?

A

Well, that's probably another talk, but trust me if we automate the changes from where some control to production, we have lower risk short time to market happy and happier employees and we free up time to do work on useful stuff and what is the main purpose on a pipeline it's actually to if we treat all configuration changes as something we could possibly put into production, the main purpose of the pipeline is to take away all changes that is not fit for production.

A

So it's about testing and when you test systems, this is the way we usually do it. So we spend a lot of time doing time-consuming, expensive manual tests. So what we usually do is we build the complete system, we design it and set it up, and then we starts to do some.

A

Like manual tests after it's built, so what we should do instead is turn that pyramid upside or actually not upside down the right way and instead of doing a lot of manual tests, we should spend time writing a smaller unit test that we can run automatically on on small parts or the infrastructure and from there when those tests have succeeded, we can go on and do only a few like high-level tests on the whole system.

A

And how do we do that on hardware? Well, testing on hardware is expensive because you actually need to have the hardware there. So what we need to do is to break the system down into small pieces immutable pieces that we can test individually and make it possible to redeploy each piece of the system.

A

We need to separate config and data, and we need to do unit tests on individual parts and there's been some great talks today about how to how to deploy self and containers containerized stuff and use config management for for deploy and so on and identify the parts you have to run unreal harder and keep hardware for testing for those parts. So you can actually run the hardware test on the hardware you're going to deploy on in production and if you want to improve something you have to measure it. So monitoring is important.

A

That's all I had time to talk about that. What have you learned? Devops is not about developers or operators, but it's about culture and practices and that can be applied everywhere, even if we are only operators or infrastructure engineers or whatever, and most important make it easy to do changes in production in a safe way by implementing pipelines and automated testing, and why are we working with the infrastructure because building stuff is fun. So that's what I was going to say. Thank you.