DEPRECATED - Migrated to https://github.com/mozilla/fxa
Перейти к файлу
Phil Booth dba93f0dbf
refactor(settings): promote Env to an enum from a wrapped string
2018-11-26 12:16:14 +00:00
.circleci fix(ci): only build deployment artifacts for tags 2018-10-04 19:56:07 +01:00
config chore(config): change aws region defaults for dev and prod (#154) r=@vladikoff,@philbooth 2018-08-02 16:41:28 -04:00
scripts chore(package): create an automated release script 2018-11-20 19:36:00 +00:00
src refactor(settings): promote Env to an enum from a wrapped string 2018-11-26 12:16:14 +00:00
.gitignore feat(db): check the emailBounces table before sending email 2018-05-16 05:25:08 +01:00
.rustfmt.toml chore(code): cargo fmt 2018-10-29 21:41:23 +00:00
.travis.yml fix(ci): enable branch builds in travis 2018-11-15 07:33:23 +00:00
API.md refactor(bounces): pull all bounce/complaint code into one module 2018-10-16 07:06:12 +01:00
CHANGELOG.md chore(package): create an automated release script 2018-11-20 19:36:00 +00:00
Cargo.lock chore(package): bump version, update changelog 2018-11-20 14:23:13 +00:00
Cargo.toml chore(build): lower the debug level to decrease binary size 2018-11-22 07:34:50 +00:00
Dockerfile chore: utilize a rust-toolchain file 2018-09-17 15:56:03 -07:00
LICENSE chore(project): add the MPL 2018-05-04 07:14:43 +01:00
README.md feat(docs): mention command-line shortcuts for running binaries 2018-10-05 13:46:04 +01:00
RELEASE.md chore(package): create an automated release script 2018-11-20 19:36:00 +00:00
r feat(docs): generate developer docs with rustdoc 2018-07-04 20:53:21 +01:00
rq chore(queues): split out a separate process for queue-handling 2018-05-29 09:01:54 +01:00
rust-toolchain chore(deploy): update rust version 2018-11-09 16:00:57 +00:00
t chore(queues): split out a separate process for queue-handling 2018-05-29 09:01:54 +01:00
tdb chore(queues): split out a separate process for queue-handling 2018-05-29 09:01:54 +01:00
version.json feat(project): create healthcheck endpoints and make $PORT an env variable (#134) r=@philbooth,@vladikoff 2018-07-18 12:37:30 -04:00

README.md

fxa_email_service

Build status CircleCI License

What's this?

The FxA team had an OKR for Q2 2018 about decoupling the auth server from SES and making it possible to send email via different providers. You can read more about that OKR in the feature doc. Subsequently, some other teams expressed an interest in depending on a standalone email service too.

This repo started as our experiment to see what a decoupled email-sending service would look like, written in Rust. It is now handling some FxA email traffic in production, and we are gradually separating it from the FxA stack to run as a standalone service in its own right.

You can find out more about the structure of the code from the developer docs.

How can I set up a dev environment?

For standalone development

We're running on the Rust nightly channel, so the easiest way to get set up is with rustup:

curl https://sh.rustup.rs -sSf | sh

If you don't want to use nightly as your default channel, you can use it just for this repo instead by installing nightly and then running rustup override in this directory:

rustup install nightly
rustup override set nightly

You will also need to install Redis. If your Redis instance is not running on the default port (6379), you will need to set the correct port in config/local.json:

{
  "redis": {
    "port": "..."
  }
}

You can also set host in the same way, if your Redis instance is not running locally.

As part of the FxA stack

If you're developing for FxA, the easiest thing to do is set up fxa-local-dev. That will install everything you need for running locally, including all of the other FxA services.

How do I run the tests?

A simple cargo t will fail because some of the tests are not threadsafe (they rely on setting environment variables that will conflict with other concurrent tests). To run the tests in a single thread instead, a shell script is provided to save you some keystrokes:

./t

That script assumes you have an instance of fxa-auth-db-mysql running locally on port 8000, which will be the case if you're running fxa-local-dev

If that's not the case, don't worry. There is another script provided to save you even more keystrokes if you're running standalone:

./tdb

That script will clone a local copy of the db repo and start it in the background. At the end of the script, the db process will be left running in the background (but subsequent runs of the script won't start extra db processes, you'll only be left with that one). If you want to kill your db process, you can find it with:

ps -ef | grep "node bin/server"

How can I send an email via SES?

You'll need to set up some config with your AWS credentials. That can be with environment variables:

  • FXA_EMAIL_AWS_KEYS_ACCESS
  • FXA_EMAIL_AWS_KEYS_SECRET

Or in config/local.json:

{
  "aws": {
    "keys": {
      "access": "...",
      "secret": "..."
    }
  }
}

config/local.json is included in .gitignore, so you don't have to worry about your keys being accidentally leaked on GitHub.

Also note that the AWS IAM limits sending to approved from addresses. Again, you can set that via environment variables:

  • FXA_EMAIL_SENDER_ADDRESS
  • FXA_EMAIL_SENDER_NAME

Or in config/local.json:

  "sender": {
    "address": "verification@latest.dev.lcip.org",
    "name": "Firefox Accounts"
  },

Once you have config set, you can start the service with:

cargo r --bin fxa_email_send

Then you can use curl to send requests:

curl \
  -d '{"to":"foo@example.com","subject":"bar","body":{"text":"baz"}}' \
  -H 'Content-Type: application/json' \
  http://localhost:8001/send

If everything is set-up correctly, you should receive email pretty much instantly.

How can I send an email via Sendgrid?

The process is broadly the same as for SES. First set your Sendgrid API key, either using the FXA_EMAIL_SENGRID_KEY environment variable or in config/local.json:

{
  "sendgrid": {
    "key": "..."
  }
}

Then start the service:

cargo r --bin fxa_email_send

Or you can use the shortcut:

./r

Then set provider to sendgrid in your request payload:

curl \
  -d '{"to":"foo@example.com","subject":"bar","body":{"text":"baz"},"provider":"sendgrid"}' \
  -H 'Content-Type: application/json' \
  http://localhost:8001/send

If everything is set-up correctly, you should receive email pretty much instantly.

How do bounce, complaint and delivery notifications work?

For consistency with the implementation in the FxA auth server, three separate SQS queues are monitored for bounce, complaint and delivery notifications. Ultimately we expect to simplify this to a single queue for all three notification types.

Messages on these queues are JSON payloads of the format described in the AWS docs and encoded in src/queues/sqs/notification/mod.rs.

When a message is received, three things happen in sequence:

  1. For bounce and complaint notification types, a bounce record is created in the auth db. Errors are fatal at this point, steps 2 and 3 will not occur if the db returns an error.

  2. The message is forwarded to the auth server via a fourth, outgoing queue. An error here is not fatal.

  3. The message is deleted from the origin queue.

Currently, both the incoming and outgoing queues happen to be SQS queues but, since that's an implementation detail, the code separates them behind more abstract Incoming and Outgoing traits.

The queue URLs and region are set via config, either using environment variables:

  • FXA_EMAIL_AWS_REGION
  • FXA_EMAIL_AWS_SQSURLS_BOUNCE
  • FXA_EMAIL_AWS_SQSURLS_COMPLAINT
  • FXA_EMAIL_AWS_SQSURLS_DELIVERY
  • FXA_EMAIL_AWS_SQSURLS_NOTIFICATION

Or in config/local.json:

{
  "aws": {
    "region": "...",
    "sqsurls": {
      "bounce": "...",
      "complaint": "...",
      "delivery": "...",
      "notification": "..."
    }
  }
}

The queue-handling code runs in a different process to the main email-sending service. You can run it locally like so:

cargo r --bin fxa_email_queues

There's also a shortcut for this:

./rq