Cleanup "use_lock_for_update_repository_storage" feature flag

Contributes to https://gitlab.com/gitlab-org/gitlab/-/issues/431198

**Problem**

It's possible to run multiple migration workers for the same
project/snippet/group simultaneously
The worker can be killed and rescheduled by Sidekiq interrupt signal.
It will leave the migration in an inconsistent state.

**Solution**

Use an exclusive lock in storage migration workers. The exclusive lease
key includes a project/snippet/group id to prevent simultaneous updates.
The key value is a Sidekiq worker jid to track the owner of the update.
This setup should handle following situations:

Worker tries to migrate a repository under existing
migration (result: job is marked as failed)
Worker started a migration but was interrupted and
rescheduled. (result: job is marked as failed, lock is released)

Changelog: changed

app/workers/concerns/update_repository_storage_worker.rb

@@ -38,36 +38,32 @@ def perform(*args)
 
     container_id ||= repository_storage_move.container_id
 
-    if Feature.enabled?(:use_lock_for_update_repository_storage)
-      # Use exclusive lock to prevent multiple storage migrations at the same time
-      #
-      # Note: instead of using a randomly generated `uuid`, we provide a worker jid value.
-      # That will allow to track a worker that requested a lease.
-      lease_key = [self.class.name.underscore, container_id].join(':')
-      exclusive_lease = Gitlab::ExclusiveLease.new(lease_key, uuid: jid, timeout: LEASE_TIMEOUT)
-      lease = exclusive_lease.try_obtain
-
-      if lease
-        begin
-          update_repository_storage(repository_storage_move)
-        ensure
-          exclusive_lease.cancel
-        end
-      else
-        # If there is an ungoing storage migration, then the current one should be marked as failed
-        repository_storage_move.do_fail!
+    # Use exclusive lock to prevent multiple storage migrations at the same time
+    #
+    # Note: instead of using a randomly generated `uuid`, we provide a worker jid value.
+    # That will allow to track a worker that requested a lease.
+    lease_key = [self.class.name.underscore, container_id].join(':')
+    exclusive_lease = Gitlab::ExclusiveLease.new(lease_key, uuid: jid, timeout: LEASE_TIMEOUT)
+    lease = exclusive_lease.try_obtain
 
-        # A special case
-        # Sidekiq can receive an interrupt signal during the processing.
-        # It kills existing workers and reschedules their jobs using the same jid.
-        # But it can cause a situation when the migration is only half complete (see https://gitlab.com/gitlab-org/gitlab/-/issues/429049#note_1635650597)
-        #
-        # Here we detect this case and release the lock.
-        uuid = Gitlab::ExclusiveLease.get_uuid(lease_key)
-        exclusive_lease.cancel if uuid == jid
+    if lease
+      begin
+        update_repository_storage(repository_storage_move)
+      ensure
+        exclusive_lease.cancel
       end
     else
-      update_repository_storage(repository_storage_move)
+      # If there is an ungoing storage migration, then the current one should be marked as failed
+      repository_storage_move.do_fail!
+
+      # A special case
+      # Sidekiq can receive an interrupt signal during the processing.
+      # It kills existing workers and reschedules their jobs using the same jid.
+      # But it can cause a situation when the migration is only half complete (see https://gitlab.com/gitlab-org/gitlab/-/issues/429049#note_1635650597)
+      #
+      # Here we detect this case and release the lock.
+      uuid = Gitlab::ExclusiveLease.get_uuid(lease_key)
+      exclusive_lease.cancel if uuid == jid
     end
   end
 

config/feature_flags/development/use_lock_for_update_repository_storage.yml

----
-name: use_lock_for_update_repository_storage
-introduced_by_url: https://gitlab.com/gitlab-org/gitlab/-/merge_requests/136169
-rollout_issue_url: https://gitlab.com/gitlab-org/gitlab/-/issues/431198
-milestone: '16.6'
-type: development
-group: group::source code
-default_enabled: false

spec/support/shared_examples/workers/update_repository_move_shared_examples.rb

@@ -96,18 +96,6 @@
               expect(repository_storage_move.reload).to be_failed
             end
           end
-
-          context 'when feature flag "use_lock_for_update_repository_storage" is disabled' do
-            before do
-              stub_feature_flags(use_lock_for_update_repository_storage: false)
-            end
-
-            it 'ignores lock and calls the update repository storage service' do
-              expect(service).to receive(:execute)
-
-              subject
-            end
-          end
         end
       end
     end
@@ -172,18 +160,6 @@
             expect(repository_storage_move.reload).to be_failed
           end
         end
-
-        context 'when feature flag "use_lock_for_update_repository_storage" is disabled' do
-          before do
-            stub_feature_flags(use_lock_for_update_repository_storage: false)
-          end
-
-          it 'ignores lock and calls the update repository storage service' do
-            expect(service).to receive(:execute)
-
-            subject
-          end
-        end
       end
     end
   end