Split processes' name migration into two

As it won't be possible to start new processes after the column
is made NOT NULL and before deploying the code that uses that column.

Author: rosa

UPGRADING.md

@@ -1,12 +1,16 @@
 # Upgrading to version 0.6.x
 
 ## New migration in 3 steps
-This version adds a new migration to the `solid_queue_processes` table. This migration adds a new column that needs to be `NOT NULL`. It will run in three steps:
-1. Add the new column, nullable
-2. Backfill existing rows that would have the column as NULL
-3. Make the column not nullable and add a new index
+This version adds two new migrations to modify the `solid_queue_processes` table. The goal of that migration is to add a new column that needs to be `NOT NULL`. This needs to be done with two migrations and the following steps to ensure it happens without downtime and with new processes being able to register just fine:
+1. Run the first migration that adds the new column, nullable
+2. Deploy the updated Solid Queue code that uses this column
+2. Run the second migration. This migration does two things:
+  - Backfill existing rows that would have the column as NULL
+  - Make the column not nullable and add a new index
 
-To install it:
+Besides, it adds another migration with no effects to the `solid_queue_recurring_tasks` table. This one can be run just fine whenever, as the column affected is not used.
+
+To install the migrations:
 ```bash
 $ bin/rails solid_queue:install:migrations
 ```
@@ -17,7 +21,7 @@ Or, if you're using a different database for Solid Queue:
 $ bin/rails solid_queue:install:migrations DATABASE=<the_name_of_your_solid_queue_db>
 ```
 
-And then just run it.
+And then follow the steps above, running first one, then deploying the code, then running the second one.
 
 ## New behaviour when workers are killed
 From this version onwards, when a worker is killed and the supervisor can detect that, it'll fail in-progress jobs claimed by that worker. For this to work correctly, you need to run the above migration and ensure you restart any supervisors you'd have. 

db/migrate/20240811173327_add_name_to_processes.rb

@@ -1,18 +1,5 @@
 class AddNameToProcesses < ActiveRecord::Migration[7.1]
-  def up
+  def change
     add_column :solid_queue_processes, :name, :string
-
-    SolidQueue::Process.find_each do |process|
-      process.name ||= [ process.kind.downcase, SecureRandom.hex(10) ].join("-")
-      process.save!
-    end
-
-    add_index :solid_queue_processes, [ :name, :supervisor_id ], unique: true
-    change_column :solid_queue_processes, :name, :string, null: false
-  end
-
-  def down
-    remove_index :solid_queue_processes, [ :name, :supervisor_id ]
-    remove_column :solid_queue_processes, :name
   end
 end

db/migrate/20240813160053_make_name_not_null.rb

@@ -0,0 +1,16 @@
+class MakeNameNotNull < ActiveRecord::Migration[7.1]
+  def up
+    SolidQueue::Process.where(name: nil).find_each do |process|
+      process.name ||= [ process.kind.downcase, SecureRandom.hex(10) ].join("-")
+      process.save!
+    end
+
+    change_column :solid_queue_processes, :name, :string, null: false
+    add_index :solid_queue_processes, [ :name, :supervisor_id ], unique: true
+  end
+
+  def down
+    remove_index :solid_queue_processes, [ :name, :supervisor_id ]
+    change_column :solid_queue_processes, :name, :string, null: false
+  end
+end

test/integration/forked_processes_lifecycle_test.rb

@@ -104,7 +104,7 @@ class ForkedProcessesLifecycleTest < ActiveSupport::TestCase
     no_pause = enqueue_store_result_job("no pause")
     pause = enqueue_store_result_job("pause", pause: 0.2.seconds)
 
-    signal_process(@pid, :INT, wait: 0.1.second)
+    signal_process(@pid, :INT, wait: 0.3.second)
     wait_for_jobs_to_finish_for(2.second)
 
     assert_completed_job_results("no pause")