Last week I had a chance to attend Puppet Camp 2016. Puppet Camp is a one day event that is held once a year in many places around the world including Australia. This time it was the fourth Melbourne conference, which gathered 240 attendees and several key partners, such as NetApp, Diaxion and Katana1.
In this blog post I want to give a quick overview of the keynote, customer and partner sessions, as well as my key takeaways from the conference.
First Impressions
I’ve never been to Puppet Camp before and this was my first experience. Sheer number of participants clearly shows that areas of configuration management and DevOps in general attracts a lot of attention of both customers and channel.
You may have heard how Cisco in Q3 of 2015 announced Puppet support for the Nexus 3000 and 9000 series switches. This was not just an accident. I had a chance to speak to NetApp, who was one of the vendors presented at the conference, and they now have Puppet integration with their Data ONTAP / FAS platform, as well as E-Series and recently acquired SolidFire line of storage arrays. I’m sure many other hardware vendors will follow.
Keynote and Puppet Update
The conference had one track of sessions spread out throughout the day and was opened by a keynote from Robert Finn, APJ Sales Director at Puppet, who was talking about the raising complexity of modern IT environments and challenges that come with it. We have gone from tens of servers to hundreds of VMs and are now on the verge of the next evolution from hundreds of VMs to thousands of containers. We can no longer manage environments manually and that is where tools such as Puppet come into play and let us manage configuration and provisioning at scale.
Rob also mentioned the “State of DevOps Report” an annual survey Puppet has been running now for five years in a row. In 2016 they collected responses from 4600 technical professionals and shared a lot of their findings in a public report, which I’ll link in the references section below.
Key takeaways: introducing configuration management in their software development practices organizations were able to achieve 3x lower change failure rate and 24x faster recovery from failures.
Ronny Sabapathee, Puppet Solutions Engineer gave an overview of the new features in the latest Puppet Enterprise 2016.4, such as corrective change reporting, changes to Puppet Orchestrator, enhancements in Code Manager and API improvements.
Key takeaways: Puppet ecosystem is growing quickly with Docker module, Jenkins plugin, significant enhancements in Azure module and VMware vRealize Automation/Orchestrator integration coming soon.
Customer Sessions
Rob Kenefik from SpecSavers spoke about their journey of scaling free version of Puppet from 10 to 290 nodes, what issues they came across and what adjustments they had to make, especially around the DB back-end.
Key takeaways: don’t use embedded Puppet database for production deployments. PostgreSQL (which is now default) provides required scalability.
Steve Curtis from ANZ briefly discussed how they automated deployment of Application Performance Monitoring (APM) agents using Puppet. Steve also has a post in Puppet blog, which I’ll link below.
Chris Harwood from Healthdirect Australia touched on a sensitive topic of organizational silos and how teams become too focused on their own performance forgetting about the customers, who should be the key priority for businesses offering customer-facing services.
Then he showed how Healthdirect moved some of the ops people to development teams giving devs access to infrastructure and making them autonomous, which significantly improved their development workflows and release frequency.
Key takeaways: DevOps key challenges are around people and processes, not technology. Teams not collaborating and lengthy infrastructure change management processes can significantly hinder development teams’ performance.
Partner Sessions
Dinesh Siriwardhane who represented Versent compared pros and cons of master/agent vs. masterless Puppet deployments and showed a demo on Puppet certificate management.
Key takeaways: Puppet master simplifies centralized management, provides reporting capabilities, but can be a single point of failure. Agentless deployment using GitHub has no single points of failure and is free, but can have major security repercussions if Git repository is compromised.
Kieran Sweet and Pedram Sanayei from Sourced made a presentation on Puppet integration with Azure and how using Puppet instead of just the low-level Azure APIs and PowerShell, can significantly simplify deployment and configuration management in the Microsoft cloud.
Key takeaways: Azure Resource Manager is a big step forward from the old Azure Service Management (classic deployment model). In light of the significant recent enhancements in the Azure Puppet module, this can become a reasonable alternative to AWS.
Scott Coulton from Autopilot closed the conference with a session on Puppet integration with Docker and more specifically around container orchestration tools, such as Docker Swarm, Kubernetes, Mesos and Flocker. Be sure to check Scott’s blog and GitHub repository where you can find a Puppet module for Docker Swarm, Vagrant template and more.
Key takeaways: Docker can be used to deploy containers, but Puppet is still essential to keep configuration across the hosts consistent.
Conclusion
I spoke to a lot of customers at the conference and what became apparent to me was that Puppet is not just another DevOps tool amongst the many. It has a wide ecosystem of partners and has gone a long way since they started as a small start-up 12 years ago in 2005.
It has a strong use case for general configuration management in Linux environments, as well as providing application configuration consistency as part of CI/CD pipelines.
Speaking of the conference itself I was pleasantly surprised by the quality of sessions and organization in general. Puppet Camp will definitely stay on my radar. I’d love to come back next year and geek out with the DevOps crowd again.
References
Niktips's Blog 