DEVOPS WEEKLY ISSUE #430 - 24th March 2019
Configuration formats, data pipeline design, local development environments, getting started with serverless and several posts on managing multi-user Kubernetes clusters. Lots of good technical case studies and opinions to start the week.
From our sponsor, VictorOps
Reducing MTTA and MTTR takes 5 simple steps. Check out this recent blog series, Reducing MTTA, to find 5 simple steps for improving incident response, lowering MTTA over time and making on-call suck less for DevOps and IT:
http://try.victorops.com/devopsweekly/reducing-mtta-alerts
News
Over the 8 years of Devops Weekly, YAML has become more and more common for various purposes. This post defends some of that usage while pointing out problematic usage patterns. The conclusion; YAML as data format is defensible. YAML as a programming language is not.
https://blog.atomist.com/in-defense-of-yaml/
A detailed post on setting up an EKS cluster. It covers doing things manually as well as using eksctl. The post also dives into securing the cluster and some observability options.
https://www.beyondvirtual.io/blog/simplifying-EKS-deployments-and-managing-them-bs/
A useful discussion for anyone building data processing pipelines, in particular with Kafka. It looks at whether you are best writing to Kafka first or to the database first and when which option might be suitable.
https://medium.com/@gwenshapira/the-case-for-database-first-pipelines-f86240c69863
Multi-user Kubernetes clusters pose a challenge around managing who can access what. Kubernetes RBAC provides part of the solution, but you probably already have some identity service. This post explores mapping Google identities to Kubernetes RBAC to help manage access.
https://medium.com/cruise/open-sourcing-rbacsync-48758df685b0
An interesting post on setting up infrastructure in regions (in this case Russia) with data locality laws and no large cloud provider services.
https://www.cloudops.com/2019/03/to-russia-with-love-deploying-kubernetes-in-foreign-locations/
A solid argument that different software solves different problems, in this case choosing Nomad over Kubernetes when all that was needed was simple service orchestration.
https://matthias-endler.de/2019/maybe-you-dont-need-kubernetes/
An in-depth look at starting to implement serverless. Looking at the details behind several services, and some general rules and tools to help you take advantage of the ecosystem.
https://read.acloud.guru/journey-to-serverless-d3256d91af16
People invariably have strong opinions about local development environments, and what place containers have in them. This post puts forward some principles and shows an example of a Node JS based development environment.
https://medium.com/@justkrup/using-docker-in-development-ad5718db08bd
Jobs
Senior Devops Engineer @ Google, via Citrusbyte, West Coast
Citrusbyte is working with Google to build a DevOps team and has the ability to give a few individuals the opportunity to work for Google. You will work with the Google Cloud team to help onboard high-profile and strategic Google accounts to GCP. You’ll work with modern DevOps tools like Terraform, Kubernetes, and GCP and implement practices such as those from the Site Reliability Handbook.
https://citrusbyte.applytojob.com/apply/DIZpovyqjC?source=DevopsWeekly
Tools
JVM Profiler is a new project providing a powerful profiling tool for Java and other JVM environments, focused on distributed applications. The agent supports sending metrics to Kafka, the console or implementing a custom reporter,
https://eng.uber.com/jvm-profiler/
https://github.com/uber-common/jvm-profiler
Work-in-progress but useful looking. As mentioned in a post above, managing authorisation for a multi-user Kubernetes cluster can be challenging. Who-can lets you interogate who can manipulate resources on your cluster. Would make a great kubectl plugin.
https://github.com/aquasecurity/kubectl-who-can
Reducing MTTA and MTTR takes 5 simple steps. Check out this recent blog series, Reducing MTTA, to find 5 simple steps for improving incident response, lowering MTTA over time and making on-call suck less for DevOps and IT:
http://try.victorops.com/devopsweekly/reducing-mtta-alerts