Is there such a thing as overautomation?

I've been a big advocate of automating everything for a while now... But a recent project I was hacking on changed my mind.

The project is https://github.com/jayunit100/SparkStreamingCassandraDemo .

I  initially set this project up to be a POC/Snippet of ETL from Spark into cassandra.  You know, one of those one day hacks where you pull code in from a few blog posts and then make sure the integration glue all works.

what next ? 

After getting it working I thought "hey this is great ! let me set up a mini jenkins server and a vagrantfile".  So I spent a few hours on that.  Then jenkins went down.  Then I found out that actually the Vagrantfile I created was depending on a somewhat unmaintained centos7 box of my own, in vagrant cloud.  So the project very quickly became much more than a POC, and was now a maintainence nightmare - I was essentially building and maintaining my own immutable and reproducible bigdata stack. 

it gets worse.

Later on, after I had my vagrantfile working, I actually decided I wanted to containerize everything.  So the project was now going to have to change - I would have to update the Vagrantfile and also update a README to tell people to install Vagrant 1.6 and so on... And get docker containers tested and working as well in a reproducible way.

a better approach.

When im hacking on a new project, I'm not too worried about reproducibility.  I can trust myself that, if indeed the project needs some reproducible deployments, I can string together a good Dockerfile/Vagrantfile after first just getting it working in a ad hoc sort of way. 

so when to dockerize/vagrantize?

I still think there is value in doing this ridiculously early.  maybe wait one week of hacking on a project before you automate everything.

but, if your not sure : i say still err on the side of automation.

I never *really* regret automation that much.  In general, Im going to come full circle and say, there probably in the long run isn't such a thing as overautomation, but sometimes , in the short term, it sure can *seem to be* be a distraction from the core work your doing.

but *seem to be* != *is* :)