This post was written by Jovile Bartkeviciute.
Key points from Operability.io Day 2:
OPERABILITY AND EVOLVABILITY – Rebecca Parsons
- Definition of done – you cannot be 75% done, done is done.
- Evolutionary architecture supports incremental and guided change across multiple architectural dimensions.
- You want deployments to be boring – you cannot experiment with this.
- It is irresponsible to use microservises architecture if you do not have mature organisation, devops culture and continuous delivery. Regardless of the benefits, without the high level of communication, microservices puts too much weight on the operations department.
- Continuous delivery is all about making deployment repeatable.
- With the central role the technology is playing in our businesses – the operations are treated seriously in they way they weren’t previously.
- “Devops, Continuous Delivery and Evolutionary Architecture are a locked set.”
THE MORNING PAPER ON OPERABILITY – Adrian Colyer
Adrian Colyer shared his favourite papers about operations issues and their possible solutions. Some key concepts:
- Operability starts with design and development. “Most Operations issues however, either have their genesis in design and development or are best solved there.”
- The three tenets:
- Expect failures to happen regularly, and handle them gracefully
- Keep things as simple as possible
- Automate everything
- Cultural aspects of the devops transition far outweigh any technical elements.
- Why monitor & measure? “Even best engineers routinely guess wrongly on root cause of end to end problems.”
- If the development team is frequently called in the middle of the night, automation is the likely outcome. If operations is frequently called, the usual reaction is to grow the operations team.
- Complex systems run in degraded mode as their normal mode of operation.
Adrian’s blog post on Operability: https://blog.acolyer.org/2016/09/21/the-morning-paper-on-operability/
HUMAN SCALE SYSTEMS – Mark Imbriaco
- Adaptability is the unfair advantage. (for humans and software)
- “Have you ever noticed that NoOps has Oops included right in the name?”
- Fight hero culture – there will always be more problems, push back and enforce healthy balance
- DevOps has become a coping mechanism for handling rate of change rather than a cultural movement focused on collaboration. The way it is usually implemented, we may as well call it continuous delivery.
- Learn from success as well as failure. Retrospectives should be a habit, not an opportunity for assigning blame.
- Even at a very small scale, automation is now a necessity not a luxury.
- “Collaborate by default. It’s not safe to go alone”
WHAT ABOUT THE JEDI INSIDE OF YOU? – COMMUNICATION – YOUR POWERFUL TOOL! – Sabine Wojcieszak
Sabine Wojcieszak wins the prize for the best outfit! She dressed up as a jedi (!!!) and blew our minds with her great Star Wars themed talk about the light and dark side of communications.
- “Communication is like dancing on the razor’s edge. You’re on the line between light and dark”
- Email communication is like big data: you will lose a lot if you are only offering words in your transmission.
- “Your focus determines your reality. Your filters are very individual”
- If you want to change – start small, your engagement matters!. You are responsible for yourself first and only then you can be responsible for your team.
- Speak of the problems before they grow into big elephants.
- Always be aware of your own goals.
- Communication effectiveness: 7% words, 38% voice, 55% body language
- “Listen with all your senses. Your communication can be the destroying light-saber or the binding element.”
FROM REVOLUTION TO EVOLUTION: UNIKERNELS AND DOCKER – Richard Mortier
- New devices, old software, generational bugs. Most of them will never be fixed – we cannot physically locate all affected IoT devices.
- Microservices: tip of the Iceberg. No matter how simple the API consumer, the OS kernel is always a huge dependency.
- Complexity is the enemy:
- Many pieces means complex configuration
- Duplication of functionality leads to inefficiency
- Large image sizes cause long boot times
- More lines of code gives a larger attack surface
- Unikernels solve the problem – they compile your application into a specialized operating system that includes only the functionality required on the target platform.
A BRIEF HISTORY OF MONITORING – Steven Acreman
Steven Acreman shared his experience with monitoring and how it has changed over the years.
- Software has got more complex
- You can’t get away with just up and down health checks
- You need a mix of operational and development metrics
- Unified monitoring doesn’t exist so pick a few good tools ( custom, APM, logs)
- Monitoring is an investment you make every single day
- Regarding time spent in operations: 50% deployment, 50% measuring, monitoring.
- “Netflix – monitoring company, that also does films.”
AUTOMATION AND OPERABILITY: THE GOOGLE SRE PERSPECTIVE – Niall Murphy
- Relationship between automation and operability:
- The platonic piece of software is perfectly automated, meaning autonomous.
- Every place where we have missed an oportunity to provie a fluid, well-behaved operator interface, we also missed an oportunity to provide an automatable inteface or an autonomous non-inteface.
- Conclusion – better operability served by not falicitating manual ops
- Your business goals might be impossible to meet without automation.
- Value of automation for Google SRE: consistency, faster repairs, faster action, time savings, a platform (platform pushes you towards autonomous systems)
- Time savings are quite difficult to calculate – need to take into account future people who would use the automation as well.
- Incentives matter:
RUNNING IOT SERVICES, EXPERIENCE REPORT – Martin Trojer
- Everything can be IoT now:
- Home automation, sensors networks, industrial control, agriculture, surveillance, building management, smart cities, healthcare, drones.. etc.
- IoT usage patterns:
- Data from sensor networks are meant to give an insight that generates business value
- City planning (parking spaces)
- Resource optimisation (building management)
- Mixing multiple datasources
- Move from monolith to ‘smallish-services’, not microservices
- When you are on a budget, microservices are expensive
- Your company might not be ready for microservices
- Process is simpler and faster
- Less services / less ops headaches
- REST > Bus – easier architecture, lots of monitoring tools, no service discovery needed.
Overall, the talks have been fun, informative and casual. Repeating ideas were that automation and monitoring are essential, microservices are great, but not applicable in every situation, devops and continuous delivery has to come first. It is all about changing company culture, communication and collaboration.
Looking forward to next year!
>> Day 1 <<
What other people thought: