Share your development workflow

I use Zeit's Now for hosting and deploying. It's the simplest way to deploy node or static websites. And it's almost free. You don't need to worry about infrastructure.

My answer for this as an indiehacker is very different from my answer as an employee. The last thing I want to do is make things overly complex and slow myself down.

• I usually make feature branches now, but for the first 3 months, I just worked on master
• Other than Amazon's Simple Email Service, I use Digital Ocean for everything. It's cheaper, it's faster and I don't need the complexity and layers of indirection of GCP, AWS or even Heroku.
• I do dev on localhost, don't use staging and production is on DO.
• I just pulled from git on production for the first couple of months, then set up deployment via Edeliver (an Elixir-based tool) and now I use Gitlab CI/CD. CI runs tests on every push to any branch and CD deploys to my server via Edeliver. Interestingly, though Gitlab uses docker, I don't have to use it on my server (which sacrifices some of Erlang VM's more powerful features).
• Yes, I write tests but I'm not dogmatic about it!

• I use Gitlab for repos and Gitlab CI/CD to deploy to Zeit Now.
• Since it's just me I just work on one branch and have that deploying automatically on every commit to my dev environment. Then when I want to shift code into production I do a manual deploy. To lock down dev I'm using basic auth on the node server as Now doesn't have any built in functionality for securing deployments.
• For tests I've only really written E2E tests (using Cypress.io)

Following 😊

Certainly glad to help out. I've been a software engineer for almost 10 years now.

• Git: depends on the project. For side projects (i.e., projects of just me), git is basically my backup. I will commit constantly (micro commits) so that HEAD always has a working bleeding edge. Stuff that doesn't work gets nuked (git reset --head). For large projects, i'll do the same locally but the main difference is that i'll use branches to work on features. Plus github reviews.
• at work, I've used everything from AWS to heroku. For small projects, I've had success with heroku (ease of use). But cost-wise, a docker-based platform is most likely more cost effective. (And it's not really that more complex) I haven't really liked AWS because of its documentation.
• in my previous job, our product was a linux binary. We used docker for everything. Development was in local docker containers. For staging (for us QA), we published the same docker containers to AWS. "Production" meant building binaries in docker and publishing to AWS.
• tests. I'm gonna start by saying that 100% coverage is a pipe dream. The reason is because tests are code coupled to the production code and it needs someone to maintain it. The ROI is often prohibitive. Instead I tend to prefer "risk" based testing---test important features, features that are complex, etc. I often prefer the inverted testing pyramid (lots of unit tests relative to integration tests). But for recent API work, I threw that rule out the window and had lots of integration tests relative to unit tests (I guess because the APIs were simple and it was more important to test "workflows" involving multiple APIs).

I use git with a master and WIP branch. I push to a remote everynight. You never know when gremlins will steal your laptop

I am aws man myself with terraform for structure. Though Im looking to do more static work with gridsome and netlify to save on costs.

Managing environments is usually something too advanced for my solo projects so I have local and prod.

I usually just have a script that runs my tests and does deploys locally with terraform. Doing a complete throwaway and rebuild of everything except data in a db.

I write tests in jest as most my work is in JS. Should probably do more cypress testing of e2e features but I just click through things most of the time.

Here's mine:

• git: I try to follow this workflow:
  https://nvie.com/posts/a-successful-git-branching-model/
  but as a solo developer I'm not sure whether it's useful, seems more suited for team work. I've never really had cases where I thought "thank goodness I'm using this workflow", so not sure I should continue on this route, and in fact I never felt i really need git O.o (!)
• I use GCP on Compute Engine, never tried Google App Engine, briefly tried Heroku when I started to learning programming but found it very opaque and couldn't understand what I was doing. I decided I should learn setting up a server from scratch so moved to GCP. I've only ever needed 1 instance since the apps I've built have served a small number of users.
• My dev workflow is as follows:

1. Build docker images locally of each service I use: Django app on a uWSGI server, postgres, redis, celery, rabbitmq, nginx
2. push images to Google Cloud Registry using gcloud cli
3. push a few simple deployment scripts to the GCP compute instance using gcloud cli
4. ssh into GCP server, pull the images and deploy with docker-compose up. Static files get uploaded into S3 with one of the deployment scripts.

For staging, nginx is configured to route requests to staging.mydomain.com to another uWSGI docker container which is configured to have staging environment variables, alongside yet another docker container for a staging postgres db. As I said, everything is on one VM instance.

I've read about tools like Travis, CI/CD etc but wasn't sure if it's worth the effort for the scope of apps / scale I'm at. Also, I don't write any code for testing.

Based on the above how can I improve my workflow?