0.2.0 • Published 9 months ago

@ryus08/penguinserver v0.2.0

Weekly downloads
-
License
BSD-2-Clause
Repository
github
Last release
9 months ago

Penguin

A server which aggregates and caches git-provider (currently just Gitlab) information

Running

This app runs on Nodejs 16.x. If you have that installed, npm install will download dependencies and npm start will start the development server.

Running with docker

A docker image is published to ryus08/penguin-alarm-server. You can run this with docker run ryus08/penguin-alarm-server.

Configuring

This app uses the node-config library for configuration. You can set the environment variables specified in ./config/custom-environment-variables.yaml to set the config values those correlate to. You can also add your own config files, such as local.yaml if you are running locally and don't want your config (especially secrets) checked in. If something doesn't have an environment variable override in custom-environment-variables.yaml, this is also a good way to configure your app. For example, if you have a config file called local-production.yaml (for usage when NODE_ENV=production), you could build your own docker image with this copied into the config directory with a Dockerfile as:

FROM ryus08/penguin-alarm-server 

COPY ./local-production.yaml ./config/local-production.yaml

Then when you run your new docker image with NODE_ENV=production, your config file will be picked up in addition to the defaults baked into the penguin-alarm-serverimage

TODO

P0

Significant effort

  • Use a real cache in ./src/clients/clientbuilder
  • publish docker image

Some effort

  • Durable but non-dynamo repository. Local disk would probably be good enough
  • Refresh git provider tokens

Minimal effort

  • Set rp timeouts higher
  • CORS
  • error handling, was @cimpress-technology/belterrorhandling
  • New dbs, add to dynamo

P0.1

  • Add AuthZ ACL caching, otherwise we'll make 2 git provider calls on every route
  • More gracefully skip newrelic
  • More gracefully skip AWS ML

P1

  • Load gitlab username from the user's registered token
  • More AuthZ testing, even if just manual
  • Use a representative user's token instead of configured god token in the poller. Maybe just set it up as the config "owner"
  • An authZ mechanism which uses a list of configured users who are allows to pair by email. I.e. "the OIDC provider seems to say you are person@foo.com, we'll trust that you are the same person the git provider says is person@foo.com"
  • Add AuthZ for predictions. Skipped because they are config agnostic and we didn't have a way to use it yet
  • This README. How to build, run, and deploy
  • Show reviewers
  • Strategy pattern for bitbucket and github
  • Show the scoring algorithm. Add exclude list of users
  • don't use app.locals.gitLabclient in any routes. Just for polling
  • Upload a token instead of using oauth
  • Notifications from gitprovider on need to refresh instead of polling
  • Anything commented out in the config. These are nice to haves
  • Get ML working without AWS. And write down how to get it working with AWS
  • Get newrelic working behind an interface for other APM solutions
  • Get the cimpress.yaml config in a better place for reintegrating for Cimpress.
    • Loading newrelic monitoring into the app and poller, see comment in app.js and pollers/poll.js
  • Upgrade dependencies.
  • Smoke tests
  • Get the poller to trigger immediately on a new config
  • Add new preferences methods to customizr
  • req/res logging