We use AWS for our infrastructure, but not the AWS Alteryx Product.
We use teraform to build up and tear down servers. We use an AMI with the proper installation. We plan to rebuild the AMI twice a year for upgrades so some worker changes are done manually and tracked with a wiki. When we rebuild the AMI we back in some or all of these latest worker changes.
If you've heard of the Cattle vs. Pets concept we aren't to Cattle yet, more like Cattle as pets... We can't just go spin down and spin up workers or controllers like you might on a traditional web app. We have post spin up steps that require manual intervention.
I was somewhat skeptical of our approach early in our process. I've only had Pets before now and we spent a great deal of time perfecting the teraform... but now that we have it we can spin up a new environment in a couple hours. We still have ongoing costs of managing the AMIs and such, but its working well so far for us.
We also came up with an alternative for HA. We didn't go with recommended clustering, we went with a warm standby running an HA license (we're controller + Gallery with separate workers)... We leave the services on the HA box off, but otherwise the server is ready. We have a manual failover procedure that takes about 5 minutes after we decide to fail over (stop the primary service if not already down, clear the locks in mongo, start the HA services).
Let me know if you have questions and we can try to share what we know.
