Merge branch '5928-geo-rake-task-to-force-housekeeping-on-next-sync' into 'master'

Resolve "Geo: Rake task to force housekeeping on next sync"

Closes #5928

See merge request gitlab-org/gitlab-ee!5623

Gemfile

@@ -184,6 +184,9 @@ gem 'httparty', '~> 0.13.3'
 # Colored output to console
 gem 'rainbow', '~> 2.2'
 
+# Progress bar
+gem 'ruby-progressbar'
+
 # GitLab settings
 gem 'settingslogic', '~> 2.0.9'
 

Gemfile.lock

@@ -1187,6 +1187,7 @@ DEPENDENCIES
   rubocop-rspec (~> 1.22.1)
   ruby-fogbugz (~> 0.2.1)
   ruby-prof (~> 0.17.0)
+  ruby-progressbar
   ruby_parser (~> 3.8)
   rufus-scheduler (~> 3.4)
   rugged (~> 0.27)

doc/administration/raketasks/geo.md

+# Geo Rake Tasks
+
+## Git housekeeping
+
+There are few tasks you can run to schedule a git housekeeping to start at the 
+next repository sync in a **Secondary node**:
+
+### Incremental Repack
+
+This is equivalent of running `git repack -d` on a _bare_ repository.
+
+**Omnibus Installation**
+
+```
+sudo gitlab-rake geo:git:housekeeping:incremental_repack
+```
+
+**Source Installation**
+
+```bash
+sudo -u git -H bundle exec rake geo:git:housekeeping:incremental_repack RAILS_ENV=production
+```
+
+### Full Repack
+
+This is equivalent of running `git repack -d -A --pack-kept-objects` on a 
+_bare_ repository which will optionally, write a reachability bitmap index
+when this is enabled in GitLab.
+
+**Omnibus Installation**
+
+```
+sudo gitlab-rake geo:git:housekeeping:full_repack
+```
+
+**Source Installation**
+
+```bash
+sudo -u git -H bundle exec rake geo:git:housekeeping:full_repack RAILS_ENV=production
+```
+
+### GC
+
+This is equivalent of running `git gc` on a _bare_ repository, optionally writing
+a reachability bitmap index when this is enabled in GitLab.
+
+**Omnibus Installation**
+
+```
+sudo gitlab-rake geo:git:housekeeping:gc
+```
+
+**Source Installation**
+
+```bash
+sudo -u git -H bundle exec rake geo:git:housekeeping:gc RAILS_ENV=production
+```

ee/app/models/geo/project_registry.rb

@@ -72,10 +72,16 @@ class Geo::ProjectRegistry < Geo::BaseRegistry
     Gitlab::Redis::SharedState.with { |redis| redis.del(fetches_since_gc_redis_key) }
   end
 
+  def set_syncs_since_gc!(value)
+    return false if !value.is_a?(Integer) || value < 0
+
+    Gitlab::Redis::SharedState.with { |redis| redis.set(fetches_since_gc_redis_key, value) }
+  end
+
   private
 
   def fetches_since_gc_redis_key
-    "projects/#{project.id}/fetches_since_gc"
+    "projects/#{project_id}/fetches_since_gc"
   end
 
   def never_synced_repository?

ee/changelogs/unreleased/5928-geo-rake-task-to-force-housekeeping-on-next-sync.yml

+---
+title: 'Geo: Rake task to force housekeeping on next sync'
+merge_request: 5623
+author:
+type: added

ee/lib/tasks/geo/git.rake

+namespace :geo do
+  namespace :git do
+    namespace :housekeeping do
+      using ProgressBar::Refinements::Enumerator
+
+      desc "GitLab | Git | Housekeeping | Garbage Collection"
+      task gc: :gitlab_environment do
+        flag_for_housekeeping(Gitlab::CurrentSettings.housekeeping_gc_period)
+      end
+
+      desc "GitLab | Git | Housekeeping | Full Repack"
+      task full_repack: :gitlab_environment do
+        flag_for_housekeeping(Gitlab::CurrentSettings.housekeeping_full_repack_period)
+      end
+
+      desc "GitLab | Git | Housekeeping | Incremental Repack"
+      task incremental_repack: :gitlab_environment do
+        flag_for_housekeeping(Gitlab::CurrentSettings.housekeeping_incremental_repack_period)
+      end
+
+      def flag_for_housekeeping(period)
+        Geo::ProjectRegistry.select(:id, :project_id).find_in_batches.with_progressbar(format: '%t: |%B| %p%% %e') do |batches|
+          batches.each do |registry|
+            registry.set_syncs_since_gc!(period - 1)
+          end
+        end
+      end
+    end
+  end
+end

ee/spec/tasks/geo/git_rake_spec.rb

+require 'rake_helper'
+
+describe 'geo:git:housekeeping' do
+  set(:project) { create(:project, :repository) }
+  set(:registry) { ::Geo::ProjectRegistry.find_or_create_by!(project: project) }
+
+  shared_examples 'housekeeping task' do |task_name, period_name|
+    it "sets existing projects syncs_gc count to #{period_name}-1" do
+      period = Gitlab::CurrentSettings.send(period_name)
+
+      expect { run_rake_task(task_name) }.to change { registry.syncs_since_gc }.to(period - 1)
+    end
+  end
+
+  before do
+    Rake.application.rake_require 'tasks/geo/git'
+    silence_progress_bar
+  end
+
+  after do
+    registry.reset_syncs_since_gc!
+  end
+
+  describe 'geo:git:housekeeping:gc' do
+    it_behaves_like 'housekeeping task', 'geo:git:housekeeping:gc', :housekeeping_gc_period
+  end
+
+  describe 'geo:git:housekeeping:full_repack' do
+    it_behaves_like 'housekeeping task', 'geo:git:housekeeping:full_repack', :housekeeping_full_repack_period
+  end
+
+  describe 'geo:git:housekeeping:incremental_repack' do
+    it_behaves_like 'housekeeping task', 'geo:git:housekeeping:incremental_repack', :housekeeping_incremental_repack_period
+  end
+end

spec/support/helpers/rake_helpers.rb

@@ -13,6 +13,10 @@ module RakeHelpers
     allow(main_object).to receive(:print)
   end
 
+  def silence_progress_bar
+    allow_any_instance_of(ProgressBar::Output).to receive(:stream).and_return(double().as_null_object)
+  end
+
   def main_object
     @main_object ||= TOPLEVEL_BINDING.eval('self')
   end