Optimising build time

[dragarcia]

With the introduction of plv8, build times for new images have gone up to unfavourable timings. Biggest factor thus far is RAM allocated to the image upon building.

Digital Ocean

• Increasing RAM at build time is not ideal as Digital Ocean plans are tagged to both RAM and disk space.
• Going for a plan with high RAM equates to higher disk space as well.
• We would be unable to scale down the image afterwards

Possible Solutions

• Build locally then deploy. (Perhaps with something like Vagrant)
• Build on top of an existing image with v8 already installed.
• Build on top of previous version.

AWS

• It is possible to increase RAM during build time and scale it accordingly in both directions for instances that would be produced from the resultant AMI.

Docker

• Not that of a worry as building it is done locally first then eventually deployed on to Docker Hub.

[dragarcia]

tl;dr - We build on top of an existing version of Supabase Postgres (We take the one from the marketplace) instead of from scratch and use Ansible tags to run specific tasks instead of everything. A guide on how to build on top can be found here.

[dragarcia]

Will close this issue once I create the same configuration for AWS (now that we have Supabase Marketplace available in their marketplace).