orgcat

A little buggy web server + DB application to test out hosting services. The goal is to be able to test hosting services for ease of (re)deployment, monitoring and debugging and more.

Quick start

To run locally in a fresh dev environment:

install .NET 7
install docker

# Dev:
./db_reset.sh                           # wipes & recreates database
dotnet run --project src/orgcat.web     # runs the web server

# see what's in the db
docker exec -it orgcat_pg psql -U postgres -d orgcat


# Run locally in a production-like mode in docker
cd src
docker-compose build && docker-compose up -d
cd orgcat.postgresdb
docker cp seeddb.sql src-db-1:/seeddb.sql
MSYS_NO_PATHCONV=1 docker exec src-db-1 psql -U postgres -d orgcat -f /seeddb.sql
# goto localhost:5057


# Deployment:
# See readmes under infra/ for deployment options.


# Other stuff:
# to reproduce the concurrency bug that makes a new survey id unusable:
seq 1 2 | xargs -n1 -P3 curl localhost:5056/survey/start/<some_new_id>

The app

functional requirements

users
- land on page ~/survey/start/<some random ascii, 7 chars>
- this starts a survey using the random ascii as the survey id
  - this is an intentional bad practice (create data on GET)
- user answers survey questions
- once questions are answered, they get a thank you message
- they can continue the survey at any time
- they can only do the survey once
admin
- LATER: can view survey results at any time
  - no auth needed, anyone can view admin page

non functional

survey should initially load in < 1s
survey pages should transition in < 1s
contains a concurrency bug so that concurrent landing on a new survey id causes an error that makes that survey id unusable

Infra requirements

at least know how to do these, don't have to implement them
separate staging and prod environments
CD: deploy on push main, goes to prod when tests pass
logging, metrics
fast rollback (skip full build, revert to previous working deployment)
maybe: structured logs with nice querying
maybe: tracing

Database

Migrations

If you've changed the data model in orgcat.postgresdb, create a migration:

cd src/orgcat.postgresdb
dotnet ef migrations add <name of migration>
dotnet ef database update   # applies migrations to your local database

Testing out hosting services

Deploy the site + DB, then go through the bug runsheet. Pretend you don't know what the bug is. How easy is it to restore service and fix the bug?

To do

check fly + neon infra vs requirements
set log level to warn (appsettings)
- update concbug tag with this
try other deployment options
- app
  - app runner
  - ecs + EC2
  - ECS + fargate
  - more aws? amplify?
  - more container runners?
  - lambda?
  - EC2 (maybe later, not a likely contender)
- db
  - supabase
  - fly postgres
  - fly litefs (replicated sqlite)
  - aurora serverless
  - RDS

Maybe/later

assess observability options
- add structured logging
- add tracing?
- add asp http metrics (still in prom package?)
- do any services give you errors by endpoint out of the box? akita?
fuzz & load test via http
assess CICD tools: buildkite, github actions, aws, more?

uozuAho/orgcat