Respond to DB health in background migrations

This changes the BackgroundMigration worker so it checks for the health
of the DB before performing a background migration. This in turn allows
us to reduce the minimum interval, without having to worry about blowing
things up if we schedule too many migrations.

In this setup, the BackgroundMigration worker will reschedule jobs as
long as the database is considered to be in an unhealthy state. Once the
database has recovered, the migration can be performed.

To determine if the database is in a healthy state, we look at the
replication lag of any replication slots defined on the primary. If the
lag is deemed to great (100 MB by default) for too many slots, the
migration is rescheduled for a later point in time.

The health checking code is hidden behind a feature flag, allowing us to
disable it if necessary.

1 files changed, 52 insertions, 0 deletions

diff --git a/spec/workers/background_migration_worker_spec.rb b/spec/workers/background_migration_worker_spec.rb
index d67e7698635..3bd072e7125 100644
--- a/spec/workers/background_migration_worker_spec.rb
+++ b/spec/workers/background_migration_worker_spec.rb
@@ -3,6 +3,12 @@ require 'spec_helper'
 describe BackgroundMigrationWorker, :sidekiq, :clean_gitlab_redis_shared_state do
   let(:worker) { described_class.new }
 
+  describe '.minimum_interval' do
+    it 'returns 2 minutes' do
+      expect(described_class.minimum_interval).to eq(2.minutes.to_i)
+    end
+  end
+
   describe '.perform' do
     it 'performs a background migration' do
       expect(Gitlab::BackgroundMigration)
@@ -28,5 +34,51 @@ describe BackgroundMigrationWorker, :sidekiq, :clean_gitlab_redis_shared_state d
 
       worker.perform('Foo', [10, 20])
     end
+
+    it 'reschedules a migration if the database is not healthy' do
+      allow(worker)
+        .to receive(:always_perform?)
+        .and_return(false)
+
+      allow(worker)
+        .to receive(:healthy_database?)
+        .and_return(false)
+
+      expect(described_class)
+        .to receive(:perform_in)
+        .with(a_kind_of(Numeric), 'Foo', [10, 20])
+
+      worker.perform('Foo', [10, 20])
+    end
+  end
+
+  describe '#healthy_database?' do
+    context 'using MySQL', :mysql do
+      it 'returns true' do
+        expect(worker.healthy_database?).to eq(true)
+      end
+    end
+
+    context 'using PostgreSQL', :postgresql do
+      context 'when replication lag is too great' do
+        it 'returns false' do
+          allow(Postgresql::ReplicationSlot)
+            .to receive(:lag_too_great?)
+            .and_return(true)
+
+          expect(worker.healthy_database?).to eq(false)
+        end
+      end
+
+      context 'when replication lag is small enough' do
+        it 'returns true' do
+          allow(Postgresql::ReplicationSlot)
+            .to receive(:lag_too_great?)
+            .and_return(false)
+
+          expect(worker.healthy_database?).to eq(true)
+        end
+      end
+    end
   end
 end