DevOpsDays Barcelona 2013 Footnotes

It was my first DevOpsDays and a fairly interesting experience.

First of all the location was impressive. Fábrica Moritz (and old brewery) in Barcelona provided an outstanding setting for the event. Then the relaxed atmosphere and people, helped to create an environment very conductive to knowledge sharing, discussion and common enlightenment.

 

Continuous Delivery

One of the topics I was very interested in (as it is something that we definitely need to me moving on at Cloudways) was Continuous Delivery .

There was a very interesting talk on the topic by Isa Vilacides from wuaki.tv. She focused on the aspects that make it difficult to implement a proper Continuous Delivery process (from manual testing, lack of automation, bad communication ...). The takeaway was that, to implement a functional Continuous Delivery system, you need first to change your business culture and mindset to one that fits into the system.

In this sense, the talk of  Francesc Bernal from venca.es, explaining the transition of a typical retail company to a DevOps culture (one of the steps needed to get to Continuous Delivery) was a very good example of the changes needed and hurdles faced when attempting this road.

New Tools for Monitoring

Another of the topics I was there for was Monitoring.  I wanted to know which are the latests trends to build highly scalable monitoring systems. We are quite happy with Zabbix at the moment. but as we grow and much more with the immediate release of our new Click&GO platform, we feel that we need something that can scale better than a MySQL based tool.

Luckily enough, there was an Open Space session about the topic with some very knowledgeable guys (sorry that I don't have the names).

The session focused a lot on Sensu and Graphite as the tools of the moment for this purpose. Several ways of using them together were discussed and also possible bottlenecks in certain configurations.

Main message I got was that, as opposed to Zabbix that is a tool that does most of the stuff you can expect from a monitoring software - agents to capture data, graphing, alerting, proactive actions ... -, these new tools follow a more strict Linux tool-kit philosophy, so - one tool for one thing -. This means you need to pair i.e. Sensu with other complementary tools to achieve the goal of a comprehensive monitoring solution that adapts to your specific needs. Afaik, I assume this to mean a steeper learning curve! :)

Kanban for Systeams

Another of the interesting Open Space sessions covered how to organize systeams in an agile/lean manner. Focus was on Scrum and Kanban.

Surprisingly, not many of the attendants to the session had a clear view on how systeams were organized within their businesses. Discussion was centred around the flexibility needed for the type of work systeams do (i.e. the random work you can't foresee due to nature of operations). 

In this sense, at Cloudways and also venca.es people, apply a Kanban based approach to systeam workload management. There was agreement on the high level of visibility that this brings to all layers of business about the work that systeams do. Also Kanban framework itself it quite flexible (much more than Scrum in my point of view) and is much more fitted as one of the cornerstones of a Continuous Delivery system (along with many others, of course).

Scrum was seen as too rigid and imposing too many roles and defined steps for the modern systeam work.  

As a personal opinion, I think that Kanban deployed throughout all the layers of the business (dev, ops, marketing, sales ...) and with feedback from team members on how to improve the system seamlessly integrated into it, is the best way to approach the Continuous Delivery dream.

Other Bits

There was much more said and seen that I can't cover here. Some bits: 

  • Automated Testing (Jenkins, Vagrant ...) as another must for Continuous Delivery
  • Modern organizations must mix profiles / knowledge to strengthen the DevOps culture. 
  • Introducing Poka-yoke mentality while creating your systems (devise failure-proof processes). This seems to mix quite well with Kanban and surely worth exploring!.
  • Automate SOPs. Standard processes that can't be enforced are worth nothing. 

Snapshot of DevOpsDays Barcelona 2013

As a final snapshot of these two great days I wanted to share the very good DevOps definition that  Tom Levey from Appdynamics did during his presentation: 

Best DevOps definition! 

Best DevOps definition! 

Pretty sure you can meet me in more DevOpsDays