❮ zur Übersicht

synyx at the OpenSource Datacenter Conference 2016 #OSDC

  • Autor: kuehn
  • Datum: 2.05.2016

Last week we attended the Open Source Datacenter Conference #OSDC 2016 in Berlin.

It offered great presentations about open source tools in relation to devops, automation, monitoring, communication, logging, continuous delivery and more.

I especially liked that the speakers felt like attendees themselves, with all of them being happy to answer tons of questions and openly discuss their (and other’s) topics and presentations, creating a great atmosphere that felt like working with colleagues that have to solve the same issues and suffer the same pain :-)

Some detail of the most interesting conference talks I heard and what I have taken from them:

Dawn Foster (@geekygirldawn) opened up the conference with her keynote on open source culture and ways to participate, contribute and live it (AND pay the bills:-). It gave some good insight into what and who powers the open source ecosystem and why the concept works and develops.

Kris Buytaert (@krisbuytaert) talked about “Another 7 tools for your #devops stack”. Among some docker “love” we learned that (obviously) everyone else in operations suffers from DNS, too. We were quite happy to hear that we already include most of the tools Kris mentioned in our stack, like Grafana and Rundeck. Obviously it was very interesting learn about other peoples views and experience with them. We also found some new fun tools to try and evaluate in the very near future, like Consul and Jitsi.

Jonathan Boulle (@jonboulle) presented the latest evolution in Rocket and container standards. He showed a nice introduction to CoreOS, but the more interesting thing was his summary about appc and how the CoreOS team works with Docker to create an industry standard for container management. Although I had not used rkt or docker before, I found the talk very interesting and understandable from an operations perspective.

Michael Prokop (@mikagrml) gave a talk about “Continuous Integration in Data Centers”, comparing to his talk for #OSDC 2013. The most important part I took from this talk was that you can’t have enough tests for everything, even your own testing-stack and your whole CI chain. Mika also put great emphasis on codereview, not only to prevent mistakes, but most importantly to share knowledge throughout your team and company. In a side note he mentioned goss, a very lightweight testing tool we will definitely evaluate soon.

Martin Alfke (@tuxmea) talked about “ChatOps”. I was really looking forward to this presentation, because not only do we emphasize a lot on chatops in the synyx admin team, but a coworker and I are about to give a chatops-talk ourselves in a couple of weeks. It was very interesting to see how Martin had prepared his talk, focusing mostly on the relation between communication and collaboration. Following this talk i entered some great discussion about chatops, visibility and noise in chats in general. We actually had this problem figured out (worksforus) before, so we cleaned up our own hubot-interaction logger on the way home and published it the day after.

Pere Urbon-Bayes (@purbon) showed some ELKstack features in his talk about “Ingesting Logs with Style”. The whole talk was basically one big live-demo about analyzing his hobby, running. Starting with his determination not to give any cloud provider access to his personal training routines, data and routes he showed how to import his running logs into ElasticSearch, then analyze and graph it with Kibana. This was quite an eye-opener showing the power of modern tools, but also the amount of information you might upload to the vendor of your fitness wearable.

Colin Charles (@bytebot) gave us some insight on “Tuning Linux for your Database”. I had to spend a lot of time in the past to troubleshoot and tune slow databases, so I was quite happy about this talk, and Colin did not disappoint to show tools and approaches. The greatest eye-opener for me (in the entire conference) was his statement, that using LVM-Snapshots cause up to 40% performance loss for the first snapshot, adding another 2-3% for subsequent snapshots.

Thank you to all speakers, thank you synyx for sending me to this conference, thank you my lovely coworkers for accompanying me on the 7-hour trip and thanks to netways for organizing OSDC 2016, see you in 2017!