3 minute read

Definitely a theme of simplicity/complexity in a few posts this week, from considering testing in production to applying cynefin to devops to discussions of private cloud.

From our sponsor, VictorOps

Machine learning (ML) isn’t just a buzzword anymore — it’s affecting how we communicate, shop, live and respond to critical DevOps incidents. Grab your spot for this free webinar to learn about driving success in incident management with machine learning:
http://try.victorops.com/devopsweekly/devops-machine-learning

News

It’s easy to talk about testing in production, but what does that look like for complex real-world systems? This post has lots of details around stateless systems.
https://medium.com/@copyconstruct/testing-in-production-the-hard-parts-3f06cefaf592

A very nice description of the tradeoffs we make when we build large systems out of many small microservices.
https://avdi.codes/simple-is-complex/

Related to the above discussion of simple, this post looks at apply the cynefin framework to operations work. Good discussions of environment management, deployment, chatops and leadership.
https://www.cio.com/article/3440057/moving-operations-to-simple-with-cynefin.html

The Kubernetes project is increasingly prevalent. The project journey report therefore makes interesting reading, looking at where contributions come from in terms of companies, geographies, individuals and more.
http://bit.ly/2kV02xL

A writeup from last weeks Chaos Conf in San Francisco. Lots of links and descriptions of talks. I particularly liked the chaos experiments for people systems discussion.
https://www.gremlin.com/blog/chaos-conf-2019-recap/

Linux capabilities provide a wealth of low-level functionality that’s not yet well exposed in easy-to-use abstractions. This post delves into the details which might get you thinking about use cases.
https://blog.container-solutions.com/linux-capabilities-in-practice

A good overview of the recent State of Devops report results, with a focus on some of the data regarding devops security practices.
https://diginomica.com/new-data-shows-devops-security-winning-combination

70% of IT spend is still on premise. This presentation looks at why, exploring the sunk cost fallacy, efficiencies of public cloud and why you might not want a private cloud after all.
https://www.slideshare.net/spnewman/its-a-trap-176000461

As cloud services have become more widely adopted, the economics of how things are priced has become more interesting, and more immediate. This experience report looks at one example, though it’s also interesting to think about what role familiarity with specific architectures plays in cost optimisation.
https://einaregilsson.com/serverless-15-percent-slower-and-eight-times-more-expensive/

Terminology and definitions, especially around new categories, tends to require lots of back and forth and building agreement. This post takes a run at defining the properties of observability systems around addressing unknown-unknowns.
https://www.honeycomb.io/blog/so-you-want-to-build-an-observability-tool/

A presentation with lots of diagrams, explaining a CI/CD process designed for a large organisation. Good discussion of some of the design decisions, and always nice to see rich examples.
http://bit.ly/2mKf2zl

Events

KubeCon + CloudNativeCon North America is coming up in San Diego from the 18th until the 21st of November. The schedule is packed with talks on the CNCF projects like Kubernetes, Envoy and Helm as well as case studies, community meetings and more. The code KCNADOW19 will save DevOps Weekly readers 10% off the ticket price too.
http://bit.ly/2ko9SrP

The O’Reilly Velocity Conference heads to Berlin, 4–7 November. Velocity is the best place on the planet for web ops and systems engineering professionals to get expert insight on building and maintaining cloud native systems. With 4 days of practical content on cloud native infrastructure, DevOps, Kubernetes, and more, there’s something for everyone. Passes start at €676 when you use the code DEVW20 (applies to Gold, Silver, and Bronze passes). Register today!
https://oreil.ly/99PIf

Tools

Talos is a single-purpose operating system designed to provide clusters abstractions and tooling rather than focus on individual hosts. It’s built to run Kubernetes.
https://www.talos-systems.com/
https://github.com/talos-systems

Machine learning (ML) isn’t just a buzzword anymore — it’s affecting how we communicate, shop, live and respond to critical DevOps incidents. Grab your spot for this free webinar to learn about driving success in incident management with machine learning:
http://try.victorops.com/devopsweekly/devops-machine-learning

Updated: