OpenStack is a massive complicated ecosystem. It is only going to survive by the vendors that support it and probably not what you need.
Let me begin by saying that given the proper talent pool, financial resources, and time; OpenStack probably isn’t bad for some people. For most institutions it is just wrong.
I will probably write another article that outlines all the questions that I should have asked before starting this. My assumptions were not all quite valid, but I did have a sneaking suspicion given a couple very relevant experiences.
Several people asked me about a way to do big(ish) data at an affordable price. I enjoy owning hardware and infrastructure and from a business perspective that is a fairly decent endeavor in terms of static equity. I was bored with coding. The two of those items intersected and I decided to take on the task of bringing in some data clients and see if I could beat the services of other competitors and give them a better experience.
Disclosure: This site is currently running on our openstack deployment in DTAC’s private cloud
Where there are clouds of smoke, there is probably fire. - An Indian Truism
It Took Forever
I witnessed an ambitious project slide face-first into the blue and white parking stall of the private cloud mall. Then later it was used as leverage to out every participant. I mostly attributed the breakdown to lack of time resources, OS talent and political infighting. nerds (heart) aws but more on that elsewhere. Ignoring those huge red flags I figured my sysadmin abilities would be able to figure things out and surpass them.
My response: I got sysadmin skills and no politics. I’ve never failed to raise a datacenter infrastructure.
OpenStack Hate Speech
There was plenty of hating over OpenStack. More specifically there was a lot of negative commentary about Nova, Neutron and Cinder. Basically the heart of the application.
My response: I kinda need those bells and whistles so it’s worth dealing with monolithic bloat because that’s my only option.
Nerds (heart) AWS
When there is blood in the streets, buy land. - Contrarian Invetors’ Proverb
Searching for OpenStack talent was like looking for unbiased political commentary on twitter. Just ain’t there. Techs not only didn’t know it well, but most applicants that had it on their resumes didn’t want to deploy it. Out of 50 techs I talked with about private cloud work, 49 said they could help me set up with Amazon AW$. It is entirely confusing to ask someone to help you with your private cloud and they offer to help what is essentially a competing public cloud.
A side note: Most tech professionals also said they were getting a “great deal” at Amazon. No financial officer I talked to shared this sentiment. I’m sure business math was never part of comp-sci. Food for thought.
My response: Well most nerds have near-0 business strategy. Also this indicates there is a massive gap in talent so I’d like to be here.
Babylon Fortuna ergo Enterprise-Ready
The codebase is huge. And kinda old at this point. In internet years it’s about 50 years old.
My response: It is mature enough to use. I should be ashamed of myself. I deployed IBM Websphere and should have known better.
It is Complicated
Look at the diagram. Seriously Siri WHERE IS MY CAR!?!?!
BTW this image is taken from the OpenStack “Getting Started” section.
My response: Well, shit. Guess I’ve dealt with worse. I hope there isn’t magic happening. I’ll trust the Fuel installers. Oh God Fuel! don’t fail me!
It Is Fragile
Refer to the previous diagram. With so many moving parts and given the lack of talent capable of managing it there is a serious level of risk in trusting a critical production environment on this.
My response: Keep my public cloud fail overs built and ready to deploy. I’m a cheat.
It Is Vendor Driven
I kinda knew that the growth of OpenStack was tainted by vendor intrigue. I have run into this spending a considerable amount of time working with OAuth 2.0 when the “framework” was going through proposals and the lead writers were splitting.
My response: Ick
The Final Decision
While it might work for a team like RackSpace, I don’t think it fits what we need. The cost benefit of having a few drivers and some prebuilt scripts to managing images wasn’t enough to justify the cost and risk of having a system that is so fragile.
The risk and the knowns of unknowns inside the massive spaghetti of scope creep was enough that I couldn’t justify that I wouldn’t be able to jump in and fix something within less than a day. Pretty much every code fix I had to make took more than a day. More of 3-5 days to figure out each bug. That was the last straw.
We tried other stacks and I launched a multiple node infrastructure with OpenNebula. It was refreshingly vertical and self-contained. But still felt like it wanted to follow the hypervisor slipstream of OpenStack.
There was a lot of other options but I was burned out at this point.
The focus internally is more on bare-metal and OS containers. So far so good.
- There is no free ride. (Development is an Expect 2 Work Employment)
- Swiss army knives suck for self-defense and bayonets
- Don’t expect to launch a cloud out of the box that will fit your business needs
- If you believe the marketing - baaah
- Community first, politics last