For the past couple years, I have been working as the Dev and the Ops for my team at StoryXpress. I promise you it has been an amazing learning ride but for all that matters, it has been an exhaustive one. We have been hosting our services on Azure and by no means do I think is Azure developer friendly. Yes, you heard that right!
For all the .NET developers, probably Microsoft Azure is a dream come true. It is tightly bound into the ecosystem of using .NET + Power Shell + Azure. But for somebody who is working on a service oriented architecture with myriad number of technologies, Windows Azure sucks. I have completely transformed all my infrastructure into code (yes yes, I know the new hot topic around, but not why I went ahead and did it). But despite all that, I have been frustrated about the state of deployment on Azure from a platform and technology agnostic perspective. Well, I went on to tell this to the Azure platform maintainers.
@1sanyamk yes, we can!— David Justice (@davidjustice) October 7, 2015
I was rather expecting these guys to get more serious about their documentation but unfortunately it was the wrong thing to demand. As a more responsible citizen, I should have contributed to their open-source efforts.
After all this rant against the Azure platform, I decided to embark on this journey to build the complete end-to-end solution from development to deployment which just works out-of-the-box. I’ll take you through the WHY, the WHAT and the HOW.
I do not want to reinvent the wheel because there have been mammoth giants before me, who operate at scale. The amazing folks at Google, Twitter, Facebook, WhatsApp have shown the world what it is like to operate at the amazing scale. After all this, people are so reckless as to go crazy the second these services go down.
A whole bunch of services are being used everyday, you name your purpose — logging, monitoring, deployment, continuous integration and what not. With the rising popularity of containers, new problems have come to light and you have a service that is trying to solve those as well. Heck! Just give me a service that spits out code and builds me a money-making business overnight.
As obvious, the development-to-deployment market is so fragmented, I would end up paying dozens of service providers to do my job for me. After my tryst with Ops, I must confess it is a very hard job keeping a service up and running. It is a very steep learning curve if one wants to become the full-stack developer that everybody aspires to be. Probably why such a fragmented expertise exists to solve these problems.
Almost all big technology houses have built massive internal systems to solve their problems (just look up every company’s engineering blogs), because of course, everybody says they have their own solution to their own “pseudo-unique” problems! THEY LIE. EVERYBODY HAS LARGELY THE SAME PROBLEM — Moving from development to production with minimal human interference. Everybody wants to achieve everything automagically.
Let’s build a system which works out-of-the-box to solve the problem of software delivery with existing battle-tested solutions.
The DevOps swiss-knife Service deployments Scaling of services Automated service discovery Load balancing Zero configuration updates Configuration Management You never have to care about DevOps again!
I named this project, Orchestrator. I would welcome name recommendations any day! Like any complex system, Orchestrator is a composition of frameworks and services (some existing and some to-be custom built), for the modern micro-services architecture which just works out-of-the box. To be very clear, this project aims to change the way you develop and deliver code right from day one. After my long ordeal with software delivery, especially when I have been a self-learner reading out blogs from the big guns, I have conceptualized and hardened on certain philosophies that will fit every software project, big or small.
- Always build compositional systems because the world is compositional
- Each component of the large system is complete on its own (it could be meaningless without the other though)
- Separation of concerns
- Forget all the new shiny things, stick to fundamentals, implement what you can (if you have the time of course)
- Build platform agnostic
From an engineering perspective, here is what this project aims to achieve.
- An out-of-the-box configured build system based on Jenkins and Mesos.
- An out-of-the-box availability of containerization of micro-services based on Docker with support for a reliable private registry.
- An out-of-the-box ability for service discovery using Weave without developer interference.
- An out-of-the-box ability to scale virtually painlessly via the usage of Docker and Mesos.
- An out-of-the-box ability for centralized logging using Elasticsearch for storage and Kibana for visualization.
- An out-of-the-box ability to monitor every running service, persistent or ephemeral without any developer interference.
- An out-of-the-box reactive load balancer for all ephemeral services and persistent data stores alike.
- An out-of-the-box platform agnostic deployment manager with access control over all deployed resources including virtual machines, containers, data stores etc. This is the most challenging part of this project which takes infrastructure as a code one step further.
- The simplest configurations to start with, allowing for configurability and extensibility wherever worth (allows to adjust course for the large spectrum of projects out there).
A large part of technologies that I have discussed above have been used for a while now in the industry and everyone is convinced about their success. But it could all be just another facade. To get around that limitation, I would like to invite all the readers to pour in their industry and technology experience to help build this end-to-end software delivery experience.
The choice of technologies above is highly opinionated based on a personal development expertise and experience. By no means they are hardened to exist but have largely been popular tooling for the past few years.
It is an ambitious take to build a better DevOps experience in 2016. The one we all deserve!