Anonymise users background job

This gives us a simple background job which anonymises users who have
been deactivated for more than 5 years. We introduce the sidekiq-scheduler
gem for the purpose of scheduling jobs in a cron-like manner.

An ADR justifies the inclusion of sidekiq-scheduler.

(Also worth noting that the sidekiq_scheduler_spec basically just checks for
typos in the Sidekiq schedule config, in case we ever need to tweak that
again or add additional jobs.)

CHANGELOG.md

@@ -4,6 +4,8 @@
 
 [Full changelog][unreleased]
 
+- Anonymise users background job
+
 ## Release 164 – 2025-01-30
 
 [Full changelog][164]

Gemfile

@@ -42,6 +42,7 @@ gem "wicked"
 gem "strip_attributes"
 gem "breadcrumbs_on_rails"
 gem "sprockets-rails"
+gem "sidekiq-scheduler"
 
 # Authentication
 gem "devise"

Gemfile.lock

@@ -169,6 +169,8 @@ GEM
     encryptor (3.0.0)
     erubi (1.13.1)
     erubis (2.7.0)
+    et-orbi (1.2.11)
+      tzinfo
     factory_bot (6.5.0)
       activesupport (>= 5.0.0)
     factory_bot_rails (6.4.4)
@@ -182,6 +184,9 @@ GEM
     ffi (1.17.0-x86_64-darwin)
     ffi (1.17.0-x86_64-linux)
     foreman (0.88.1)
+    fugit (1.11.1)
+      et-orbi (~> 1, >= 1.2.11)
+      raabro (~> 1.4)
     globalid (1.2.1)
       activesupport (>= 6.1)
     govuk_design_system_formbuilder (5.8.0)
@@ -332,6 +337,7 @@ GEM
       rspec-expectations (~> 3.12)
       rspec-mocks (~> 3.12)
       rspec-support (~> 3.12)
+    raabro (1.4.0)
     racc (1.8.1)
     rack (2.2.10)
     rack-attack (6.7.0)
@@ -455,6 +461,8 @@ GEM
     ruby_parser (3.19.1)
       sexp_processor (~> 4.16)
     rubyzip (2.4.1)
+    rufus-scheduler (3.9.2)
+      fugit (~> 1.1, >= 1.11.1)
     selenium-webdriver (4.28.0)
       base64 (~> 0.2)
       logger (~> 1.4)
@@ -469,6 +477,10 @@ GEM
       logger
       rack (>= 2.2.4)
       redis-client (>= 0.22.2)
+    sidekiq-scheduler (5.0.6)
+      rufus-scheduler (~> 3.2)
+      sidekiq (>= 6, < 8)
+      tilt (>= 1.4.0, < 3)
     simplecov (0.22.0)
       docile (~> 1.1)
       simplecov-html (~> 0.11)
@@ -616,6 +628,7 @@ DEPENDENCIES
   selenium-webdriver
   shoulda-matchers
   sidekiq (~> 7)
+  sidekiq-scheduler
   simplecov (~> 0.22.0)
   simplecov-lcov (~> 0.8.0)
   sprockets-rails

app/jobs/anonymise_deactivated_users_job.rb

@@ -0,0 +1,9 @@
+class AnonymiseDeactivatedUsersJob
+  include Sidekiq::Job
+
+  def perform
+    User.deactivated.where("deactivated_at < ?", 5.years.ago).each do |user|
+      AnonymiseUser.new(user:).call
+    end
+  end
+end

config/routes.rb

@@ -1,6 +1,7 @@
 # frozen_string_literal: true
 
 require "sidekiq/web"
+require "sidekiq-scheduler/web"
 
 Rails.application.routes.draw do
   devise_scope :user do

config/sidekiq.yml

@@ -2,3 +2,8 @@
 :queues:
   - default
   - mailers
+:scheduler:
+  :schedule:
+    anonymise_deactivated_users:
+      cron: '0 0 * * 0' # Every Sunday at midnight
+      class: AnonymiseDeactivatedUsersJob

doc/architecture/decisions/0040-use-sidekiq-scheduler-for-scheduled-jobs.md

@@ -0,0 +1,33 @@
+# 40. Use sidekiq-scheduler for scheduled jobs
+
+Date: 2025-02-03
+
+## Status
+
+Accepted
+
+## Context
+
+We needed a mechanism for running scheduled background jobs in order to
+anonymise users who have been inactive for more than five years. There is no
+mechanism currently in place in RODA to handle scheduled background jobs, but
+we do have Sidekiq already available to us for running jobs in the background.
+
+## Decision
+
+We have decided to use [sidekiq-scheduler][1] as a lightweight scheduled job
+solution. A small amount of research suggested that this gem offered a decent
+implementation with a standard and well-understood interface (the time-honoured
+cron syntax combined with Sidekiq's jobs (_né_ workers)). We also briefly
+explored [sidekiq-cron][2] which does almost exactly the same thing, but had
+slightly worse documentation.
+
+## Consequences
+
+With the addition of this gem, we have a small amount of configuration overhead
+and, of course, one additional dependency. Neither of these is particularly
+onerous. We now benefit from having a standardised solution for our current
+and future scheduled background job requirements.
+
+[1]: https://github.com/sidekiq-scheduler/sidekiq-scheduler
+[2]: https://github.com/sidekiq-cron/sidekiq-cron

spec/config/sidekiq_scheduler_spec.rb

@@ -0,0 +1,34 @@
+require "rails_helper"
+require "fugit"
+
+RSpec.describe "sidekiq-scheduler" do
+  sidekiq_file = File.join(Rails.root, "config", "sidekiq.yml")
+  schedule = YAML.load_file(sidekiq_file)[:scheduler][:schedule]
+
+  describe "cron syntax" do
+    schedule.each do |job_name, values|
+      cron = values["cron"]
+      it "#{job_name} has correct cron syntax" do
+        expect { Fugit.do_parse(cron) }.not_to raise_error
+      end
+    end
+  end
+
+  describe "job classes" do
+    schedule.each do |job_name, values|
+      klass = values["class"]
+      it "#{job_name} has #{klass} class in /jobs" do
+        expect { klass.constantize }.not_to raise_error
+      end
+    end
+  end
+
+  describe "job names" do
+    schedule.each do |job_name, values|
+      klass = values["class"]
+      it "#{job_name} has correct name" do
+        expect(klass.underscore).to start_with(job_name)
+      end
+    end
+  end
+end

spec/jobs/anonymise_deactivated_users_job_spec.rb

@@ -0,0 +1,56 @@
+require "rails_helper"
+
+RSpec.describe AnonymiseDeactivatedUsersJob, type: :job do
+  describe "the job" do
+    subject(:job) { AnonymiseDeactivatedUsersJob.perform_async }
+
+    it "is enqueued" do
+      expect { job }.to change(AnonymiseDeactivatedUsersJob.jobs, :size).by(1)
+    end
+
+    it "is drained" do
+      job
+      AnonymiseDeactivatedUsersJob.drain
+      expect(AnonymiseDeactivatedUsersJob.jobs.size).to eq 0
+    end
+  end
+
+  describe "#perform" do
+    let(:the_recent_past) { 2.years.ago }
+    let(:the_distant_past) { 6.years.ago }
+
+    it "anonymises a user who has been inactive for more than 5 years" do
+      create(:beis_user, deactivated_at: the_distant_past)
+
+      described_class.new.perform
+
+      expect(User.first.anonymised_at).not_to eq nil
+    end
+
+    it "does not anonymise a user who has been inactive for less than 5 years" do
+      create(:beis_user, deactivated_at: the_recent_past)
+
+      described_class.new.perform
+
+      expect(User.first.anonymised_at).to eq nil
+    end
+
+    it "anonymises a set of users who were deactivated in the distant past" do
+      5.times { create(:beis_user, deactivated_at: the_distant_past) }
+
+      described_class.new.perform
+
+      expect(User.deactivated.count).to eq 0
+      expect(User.where.not(anonymised_at: nil).count).to eq 5
+    end
+
+    it "does not anonymise a set of users who were deactivated in the recent past" do
+      5.times { create(:beis_user, deactivated_at: the_recent_past) }
+
+      described_class.new.perform
+
+      expect(User.deactivated.count).to eq 5
+      expect(User.where.not(anonymised_at: nil).count).to eq 0
+    end
+  end
+end

spec/support/sidekiq.rb

@@ -0,0 +1,2 @@
+require "sidekiq/testing"
+Sidekiq::Testing.fake! # qv. https://github.com/sidekiq/sidekiq/wiki/Testing#testing-worker-queueing-fake