a259c4508e
This is the initial commit for PostgreSQL service. It contains bare minimums to introduce PostgreSQL service. Currently, we only create a virtual machine, install PostgreSQL binaries in it and do some basic configurations. In the upcoming commits we will add things like UI, API support, HA, backup/restore, firewall rules, DNS names, scaling up/down, certificates and many more. Even though this is incomplete, it provides enough infrastructure to build on top of it. Here are few notable design decisions we made in this commit; - We don't ask for database name, username or password from user. They will all be created by default. Database name and username is postgres, password will be auto generated at run-time. - Corollary of above, we are giving superuser access to users. This might be biggest divergence of our postgres service implementation from the incumbents but it is not an unseen practice. There are numerous benefits of this; most importantly, it gives more power to users, which could tilt decision to our favor for users who want or need superuser access. Secondly, it makes our implementation simpler; we don't need to implement special logic to support role or extension creation. Users can do these things themselves without needing us to provide special API/UI. On the flip side, it is possible to shoot yourself on the foot with too much power, and thus create additional support burden. I expect this to be rare occasion, but we might want to consider our support model for such cases. - We are also introducing a new concept, that is not just about postgres in this commit; service projects. PostgreSQL service only exposes PostgreSQL resource to user not VMs or public IP addresses it creates in the back end. Those resources needs to live in somewhere. Service project is exactly that. Each service will have their own account and project where they can keep the backing resources. In this commit, we are creating one project for all PostgreSQL services' resources, but after gathering more data, we might change that in the future. Currently there is no API or GUI interface. To be able to create PostgreSQL resources, you need to set ENV["POSTGRES_SERVICE_PROJECT_ID"] in your .env.rb. Then you can create new postgres database with this command; Prog::Postgres::PostgresNexus.assemble(<project-id>, <location>, "test-pg-server", "standard-4", 100) After it is ready, you can get the connection string like this; Postgres::PostgresServer.first.connection_string
10 lines
299 B
Ruby
Executable File
10 lines
299 B
Ruby
Executable File
#!/bin/env ruby
|
|
# frozen_string_literal: true
|
|
|
|
require_relative "../../common/lib/util"
|
|
|
|
# We don't change the user to postgres here. Instead we run the below
|
|
# command as root, because the cluster is running from systemd, and can
|
|
# only be restarted as root.
|
|
r "sudo pg_ctlcluster 16 main restart"
|