Don't Look Into the Light

Published on , 1003 words, 4 minutes to read

So at a previous job I was working at, we maintained a system. This system powered a significant part of the core of how the product was actually used (as far as usage metrics reported). Over time, we had bolted something onto the side of this product to take actions based on the numbers the product was tracking.

After a few years of cycling through various people, this system was very hard to understand. Data would flow in on one end, go to an aggregation layer, then get sent to storage and another aggregation layer, and then eventually all of the metrics were calculated. This system was fairly expensive to operate and it was stressing the datastores it relied on beyond what other companies called theoretical limits. Oh, to make things even more fun; the part that makes actions based on the data was barely keeping up with what it needed to do. It was supposed to run each of the checks once a minute and was running all of them in 57 seconds.

During a planning meeting we started to complain about the state of the world and how godawful everything had become. The undocumented (and probably undocumentable) organic nature of the system had gotten out of hand. We thought we could kill two birds with one stone and wanted to subsume another product that took action based on data, as well as create a generic platform to reimplement the older action-taking layer on top of.

The rules were set, the groundwork was laid. We decided:

Those of you who have been down this road before probably have massive alarm bells going off in your head. This is one of those things that looks like a good idea on paper, can probably be passed off as a good idea to management and actually implemented; as happened here.

So we set off on our quest to write this software. The repo was created. CI was configured. The scripts were optimized to dump out code coverage as output. We strived to document everything on day 1. We took advantage of the datastore we were using. Everything was looking great.

Then the product team came in and noticed fresh meat. They soon realized that this could be a Big Thing to customers, and they wanted to get in on it as soon as possible. So we suddenly had our deadlines pushed forward and needed to get the whole thing into testing yesterday.

We set it up, set a trigger for a task, and it worked in testing. After a while of it consistently doing that with the continuous functional testing tooling, we told product it was okay to have a VERY LIMITED set of customers have at it.

That was a mistake. It fell apart the second customers touched it. We struggled to understand why. We dug into the core of the beast we had just created and managed to discover we made critical fundamental errors. The heart of the task matching code was this monstrosity of a cross join that took the other people on the team a few sheets of graph paper to break down and understand. The task execution layer worked perfectly in testing, but almost never in production.

And after a week of solid debugging (including making deals with other teams, satan, jesus and the pope to try and understand it), we had made no progress. It was almost as if there was some kind of gremlin in the code that was just randomly making things not fire if it wasn’t one of our internal users triggering it.

We had to apologize with the product team. Apparently the a lot of product team had to go on damage control as a result of this. I can only imagine the trickled-down impact this had on other projects internal to the company.

The lesson here is threefold. First, the Big Rewrite is almost a sure-fire way to ensure a project fails. Avoid that temptation. Don’t look into the light. It looks nice, it may even feel nice. Statistically speaking, it’s not nice when you get to the other side of it.

The second lesson is that making something microservices out of the gate is a terrible idea. Microservices architectures are not planned. They are an evolutionary result, not a fully anticipated feature.

Finally, don’t “design for the future”. The future hasn’t happened yet. Nobody knows how it’s going to turn out. The future is going to happen, and you can either adapt to it as it happens in the Now or fail to. Don’t make things overly modular, that leads to insane things like dynamically linking parts of an application over HTTP.

If you 'future proof' a system you build today, chances are when the future arrives the system will be unmaintainable or incomprehensible.
- John Murphy


This kind of advice is probably gonna feel like a slap to the face to a lot of people. People really put their heart into their work. It feeds egos massively. It can be very painful to have to say no to something someone is really passionate about. It can even lead to people changing their career plans depending on the person.

But this is the truth of the matter as far as I can tell. This is generally what happens during the Big Rewrite centred around Best Practices for Cloud Native software.

The most successful design decisions are wholly and utterly subjective to every kind of project you come across. What works in system A probably won’t work perfectly in system B. Everything is its own unique snowflake. Embrace this.


Facts and circumstances may have changed since publication. Please contact me before jumping to conclusions if something seems wrong or unclear.

Tags: practices, big-rewrite