poja 9.1.1

Serverless Postgres+Java hosted on Github+AWS

POJA CLI

The Python CLI for maintaining a POJA stack... or dozens of them!

General usage

1. Invoke the POJA CLI depending on the use case you want to address, see section below. We recommend prefixing your poja application names with poja-.
2. Commit changes and push them to Github.
3. Define the Github secrets for deploying into your AWS prod and preprod accounts: PROD_AWS_ACCESS_KEY_ID, PROD_AWS_SECRET_ACCESS_KEY, PREPROD_AWS_ACCESS_KEY_ID, and PREPROD_AWS_SECRET_ACCESS_KEY. If you use the same account for prod and preprod, just give the same values to the prod and preprod variables.
4. Run the CD compute action. This creates the serverless Spring Boot. The API URL is printed in the Github console.

Use cases

Create a completely new project

pip install poja
python -m poja \
  --app-name=poja-base \
  --package-full-name=com.company.base \
  --region=eu-west-3 \
  --output-dir=folder-to-be-created \

  --with-own-vpc=true \
  --ssm-sg-id=/poja/sg/id \
  --ssm-subnet1-id=/poja/subnet/private1/id \
  --ssm-subnet2-id=/poja/subnet/private2/id

Those configurations will be automatically saved in poja.yml at the end of the creation. See section "Use your own VPC" for the --with-own-vpc and --ssm-xxx-id arguments.

Upgrade an already existing project

pip install poja --upgrade
python -m poja \
  --app-name=poja-base \
  ...
  --output-dir=folder-already-created

Note the --upgrade and the --output-dir=folder-already-created flags.

The POJA configuration that was used for the previous generation is saved in poja.yml: it will be updated after the new upgrade.

Configure your database

Use --with-database=posgres|sqlite|none. In particular, none is handy if you want to use an already existing Postgres, that you will manually reference through custom Java env vars.

If you want POJA to fully manage Postgres: from creation, to operations -- scale-in, scale-out to zero, DDoS protection, regular backups -- to deletion. Then do as follows before running the CD compute action:

1. Create two entries in SSM that stores the credentials of the database that will be created. The name MUST be as follows: /<?app-name>/<?env>/db/username and /<?app-name>/<?env>/db/password where <?app-name> is any name you want and <?env> is either prod or preprod.
2. Define the Github variable PROD_DB_CLUSTER_TIMEOUT that sets the prod database cluster scaling down timeout. Note that its value must be between 300 seconds (5 minutes) and 86_400 seconds (1 day). Due to the once-per-day health check action, the (serverless) prod database will always be hot if you set it to one day.
3. Run the CD storage action. This creates the serverless Postgres. The database URL is printed in the Github console.

Use custom/additional Java deps

Just provide the argument --custom-java-deps=your-list-of-deps where your-list-of-deps contains the dependency lines that are to be added to build.gradle. Here is an example of such a file.

Once the generation finishes, your-list-of-deps will be copied at the root path of the genrated directory, under the name poja-custom-java-deps.txt. That file will come handy for future generations based on past generations.

Use custom/additional Java env vars

Similar to the Java deps section above, but with --custom-java-env-vars as argument name. Here is an example of such a file.

Use your own VPC

- For the moment, you MUST use your own managed VPC for hosting the POJA generated resources.
- That is, you MUST use --with-own-vpc=true.
- If you want to put POJA resources inside AWS managed network,
- contact us and we will tell you what needs to be done.

Create first:

• Two subnets. They MUST be private, and access Internet through a NAT Gateway. Reference their id in SSM under any name you want. If you use public subnets, then some parts of POJA might not work correctly, notably the asynchronous stack and the bucket stack.
• A security group that allows incoming HTTP traffic, allows all internal traffic, and allows all outcoming traffic. Put its id in SSM under any name you want.

Warning Remind that the NAT Gateway associated to the subnets is not serverless. Whether your POJA is used or not, the NAT Gateway will generate a fixed lower cost of around $35 per month. If you host 100 POJA in the same VPC, that makes $0.35 the fixed cost per POJA.

Then invoke Poja CLI with the following additional parameters:

--with-own=vpc=true \
--ssm-sg-id=/poja/sg/id \
--ssm-subnet1-id=/poja/subnet/private1/id \
--ssm-subnet2-id=/poja/subnet/private2/id