Run serivce discovery on load balancing configuration

What does this MR do?

This MR forces an initial run of postgres service discovery as soon as load balancing is configured. This solves the problem in #323726 (closed).

Before this MR, puma would preload the application and set up load balancing with an empty host list before forking. Any database access either pre-fork or post-fork before the associated on_worker_start callback ran service discovery would then query the database primary.

https://gitlab.com/gitlab-org/gitlab/blob/master/config/initializers/load_balancing.rb#L17-27 shows this initialization process.

This MR simply runs service discovery once when load balancing is configured so that we retrieve an initial list of hosts immediately.

Screenshots or Screencasts (strongly suggested)

How to setup and validate locally (strongly suggested)

There are a few steps required to run service discovery locally. This is the easiest way I've found.

1. Install dnsmasq to run a local dns nameserver.
   • brew install dnsmasq (or your package manager of choice)
2. Start dnsmasq on port 53 (the default)
   • sudo brew services start dnsmasq (needs sudo because of the privileged port)
3. Verify that dnsmasq is working
   • dig @localhost localhost +short should return 127.0.0.1
4. Stop your local gdk postgresql
   • gdk stop postgresql
5. Run postgresql directly so that it opens a tcp port.
   • cd your-gdk-dir/postgresql/data && pg_ctl start -D .
6. Add the following production entry to your database.yml (this problem is only reproducible with RAILS_ENV=production)

production:
  main:
    adapter: postgresql
    encoding: unicode
    database: gitlabhq_development
    host: localhost
    port: 5432
    pool: 10
    prepared_statements: false
    variables:
      statement_timeout: 120s
    load_balancing:
      discover:
        nameserver: localhost
        port: 53
        record: localhost
        record_type: A
        interval: 60
        disconnect_timeout: 120

To reproduce the problem, without this branch checked out

1. rm log/database_load_balancing.log so it is clear for the problem.
2. Stop the running rails server gdk stop rails-web
3. Start puma directly env RAILS_ENV=production bundle exec puma
4. Wait until you see the messages that workers have booted, then shut down the server with ^C
5. cat log/database_load_balancing.log and see the messages with event: "no_secondaries_available"

To verify that the problem was fixed, check out this branch, and repeat these steps. The log/database_load_balancing.log file will not be recreated because no messages will be written to it.

Does this MR meet the acceptance criteria?

Conformity

• I have included changelog trailers, or none are needed. (Does this MR need a changelog?)
• I have added/updated documentation, or it's not needed. (Is documentation required?)
• I have properly separated EE content from FOSS, or this MR is FOSS only. (Where should EE code go?)
• I have added information for database reviewers in the MR description, or it's not needed. (Does this MR have database related changes?)
• I have self-reviewed this MR per code review guidelines.
• This MR does not harm performance, or I have asked a reviewer to help assess the performance impact. (Merge request performance guidelines)
• I have followed the style guides.
• This change is backwards compatible across updates, or this does not apply.

Availability and Testing

• I have added/updated tests following the Testing Guide, or it's not needed. (Consider all test levels. See the Test Planning Process.)
• I have tested this MR in all supported browsers, or it's not needed.
• I have informed the Infrastructure department of a default or new setting change per definition of done, or it's not needed.