Akkeris Database Broker

A database broker for a variety of cloud providers and even on-prem db systems that implements the Open Service Broker 2.13 compliant database broker REST API. Depending on provider it can provision postgres, mysql, aws aurora and hypothetically oracle and mssql (although not tested) databases. It can also be ran without Akkeris, but why would you?

Providers

The broker has support for the following providers

AWS RDS Instances and Clusters
Gcloud SQL Instances and Clusters
Postgres Databases via Shared Tenant
MySQL 5.5, 5.7, 8 Databases via Shared Tenant

Features

Create your own plans
Upgrade plans
Take backups, list and restore
Database Read-Only Replicas
Extra Database Accounts (read-only, read-write, create, remove, rotate password)
Database Logs
Restart
Preprovisioning databases for speed
Generate and issue KMS keys stored on fortanix

Installing

First, set your settings, so to speak, although not required these installation instructions assume you're deploying to a dockerized environment. You'll also need to provision (manually) a postgres database so the database broker can store plans, databases and other information for itself. Once you have your settings, move on to the deploy step, then finally setup your docs/PLANS.md.

1. Settings

Note almost all of these can be set via the command line as well.

Required

DATABASE_URL - The postgres database to store its information on what databases its provisioned, this should be in the format of postgres://user:password@host:port/database?sslmode=disable or leave off sslmode=disable if ssl is supported. This will auto create the schema if its unavailable.
NAME_PREFIX - The prefix to use for all provisioned databases this should be short and help namespace databases created by the broker vs. other databases that may exist in the broker for other purposes. This is global to all of the providers configured.

AWS Provider Specific

AWS_REGION - The AWS region to provision databases in, only one aws provider and region are supported by the database broker.
AWS_VPC_SECURITY_GROUPS - The VPC security groups to automatically assign for all VPC instances, this overrides any plan settings and is recommended you set this in the environment.
AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY to an IAM role that has full access to RDS in the AWS_REGION you specified above.

Note that you can get away with not setting AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY and use EC2 IAM roles or hard coded credentials via the ~/.aws/credentials file but these are not recommended!

Google Cloud Specific

Gcloud credentials are automatically inferred by the SDK through the standard environment variables, installed credentials on the host, or via server roles. See https://cloud.google.com/docs/authentication/production for more information on injecting credentials in the app (normally set GOOGLE_APPLICATION_CREDENTIALS to the path of your credentials (in json format). Ensure the credentials used have access to SQL administration.
GCLOUD_PROJECT_ID - The google project id to use.
GCLOUD_REGION - The google region used for this broker.

Shared Postgres Provider Specific

There are no environment variables for shared postgres providers, although sensitive configuration can be set in the enivornment, see docs/PLANS.md for more information.

Shared Mysql Provider Specific

There are no environment variables for shared postgres providers, although sensitive configuration can be set in the enivornment, see docs/PLANS.md for more information.

Optional

PORT - This defaults to 8443, setting this changes the default port number to listen to http (or https) traffic on
RETRY_WEBHOOKS - (WORKER ONLY) whether outbound notifications about provisions or create bindings should be retried if they fail. This by default is false, unless you trust or know the clients hitting this broker, leave this disabled.

2. Deployment

You can deploy the image akkeris/database-broker:lastest via docker with the environment or config var settings above. If you decide you're going to build this manually and run it you'll need see the Building section below.

3. Plans

Plans can be created by modifying the database table called "plans". They provide a great way of limiting the scope, capability and offerings to whomever is using the broker. See docs/PLANS.md for more information. By default the database-broker will initially load with plans for aws and shared postgres.

4. Setup Task Worker

You'll need to deploy one or multiple (depending on your load) task workers with the same config or settings specified in Step 1. but with a different startup command, append the -background-tasks option to the service brokers startup command to put it into worker mode. You MUST have at least 1 worker.

Running

As described in the setup instructions you should have two deployments for your application, the first is the API that receives requests, the other is the tasks process. See start.sh for the API startup command, see start-background.sh for the tasks process startup command. Both of these need the above environment variables in order to run correctly.

Debugging

You can optionally pass in the startup options -logtostderr=1 -stderrthreshold 0 to enable debugging, in addition you can set GLOG_logtostderr=1 to debug via the environment. See glog for more information on enabling various levels. You can also set STACKIMPACT as an environment variable to have profiling information sent to stack impact.

Contributing and Building

dep ensure
make
./servicebroker ...

Testing

make test

To run the aws instance and cluster tests TEST_AWS_CLUSTER and TEST_AWS_INSTANCE must be set to true.
To run the shared postgres tests set TEST_SHARED_POSTGRES to true (the DATABASE_URL will be used as the shared tenant!)
To run the mysql postgres test set MYSQL_URL and TEST_SHARED_MYSQL to true.

ketiko/database-broker