DEVOPS WEEKLY ISSUE #380 - 8th April 2018

2 minute read

Posts on several core practices of continuous delivery this week, from not branching to feature flags to how to manage database migrations. Also several posts on the growth of Kubernetes as a target platform for development tooling.

Sponsor

DevOps teams need real-time monitoring, alerting, and communication tools that work well together. VictorOps integrates with 100+ tools to improve system observability and team collaboration to optimize incident response, all in one place.
http://try.victorops.com/DevOpsWeekly/integration-partners

News

As part of the Kubernetes Application Definitions working group, and the Apps Special Interest Group, we’re running a survey to find out how people are deploying applications to Kubernetes. If you’re already using Kubernetes please take 10 minutes to help us understand the current landscape of tools and practices.
https://goo.gl/Vzaua3

An excellent set of details notes from SRECon, contains both opinions of the various talks as well as all the links to slides or material covered in the presentations.
https://noidea.dog/blog/srecon-americas-2018-day-1
https://noidea.dog/blog/srecon-americas-2018-day-2
https://noidea.dog/blog/srecon-americas-2018-day-3

A well argued take on why branching makes for a poor development workflow when adopting continuous delivery practices.
http://www.davefarley.net/?p=247

As Kubernetes becomes more popular for running applications the development tooling is starting to get better too. This post explores a few of the current tools for building apps for Kubernetes.
https://blog.hasura.io/draft-vs-gitkube-vs-helm-vs-ksonnet-vs-metaparticle-vs-skaffold-f5aa9561f948

Feature flags are a powerful tool but can become unwieldy in practice. This post explores a few ways of managing flag debt; from integrating with issue trackers, expiring flags and limiting the number of flags.
https://www.split.io/blog/managing-feature-flag-debt-split/

A presentation on scaling Prometheus, in particular looking at Thanos, a system for providing a global and durable approach to distributed Prometheus clusters.
https://www.slideshare.net/BartomiejPotka/thanos-global-durable-prometheus-monitoring

If you’re managing PostgreSQL you’re probably familiar with the process of vacuuming. This post explores how monitoring a few key PostgreSQL metrics and events will help you ensure that vacuum processes are proceeding as expected.
https://www.datadoghq.com/blog/postgresql-vacuum-monitoring/

Continuous deployment of stateless applications is much easier to manage that those with data. Part of the challenge involves managing schema changes in an automatic fashion. This post explores why database deploys should be decoupled from application changes.
http://www.philandstuff.com/2018/04/04/keep-database-deploys-separate.html

A post on why trying to identify a single root cause for a failure in a complex system is limiting to trying to to prevent future failures.
http://willgallego.com/2018/04/02/no-seriously-root-cause-is-a-fallacy/

An interesting approach to implementing automated rollback for Helm deployments on Kubernetes, using metrics from ElasticSearch and Prometheus.
https://container-solutions.com/automated-rollback-helm-releases-based-logs-metrics/

Tools

Textql is a handy utility for executing SQL queries against structured text like CSV or TSV.
https://github.com/dinedal/textql

Lots of people using AWS end up with multiple AWS accounts. CIDR House Rules is an application for aggregating and centralising network information like elastic IPs, load balancers and VPC configuration.
https://github.com/trulia/cidr-house-rules
https://www.trulia.com/blog/tech/trulia-open-source/

Gitkube is a tool for building and deploying docker images on Kubernetes using a Heroku-like git push workflow.
https://gitkube.sh/

DevOps teams need real-time monitoring, alerting, and communication tools that work well together. VictorOps integrates with 100+ tools to improve system observability and team collaboration to optimize incident response, all in one place.
http://try.victorops.com/DevOpsWeekly/integration-partners

Updated: