Fix race conditions for AuthorizedProjectsWorker

There were two cases that could be problematic:

1. Because sometimes AuthorizedProjectsWorker would be scheduled in a
   transaction it was possible for a job to run/complete before a
   COMMIT; resulting in it either producing an error, or producing no
   new data.

2. When scheduling jobs the code would not wait until completion. This
   could lead to a user creating a project and then immediately trying
   to push to it. Usually this will work fine, but given enough load it
   might take a few seconds before a user has access.

The first one is problematic, the second one is mostly just annoying
(but annoying enough to warrant a solution).

This commit changes two things to deal with this:

1. Sidekiq scheduling now takes places after a COMMIT, this is ensured
   by scheduling using Rails' after_commit hook instead of doing so in
   an arbitrary method.

2. When scheduling jobs the calling thread now waits for all jobs to
   complete.

Solution 2 requires tracking of job completions. Sidekiq provides a way
to find a job by its ID, but this involves scanning over the entire
queue; something that is very in-efficient for large queues. As such a
more efficient solution is necessary. There are two main Gems that can
do this in a more efficient manner:

* sidekiq-status
* sidekiq_status

No, this is not a joke. Both Gems do a similar thing (but slightly
different), and the only difference in their name is a dash vs an
underscore. Both Gems however provide far more than just checking if a
job has been completed, and both have their problems. sidekiq-status
does not appear to be actively maintained, with the last release being
in 2015. It also has some issues during testing as API calls are not
stubbed in any way. sidekiq_status on the other hand does not appear to
be very popular, and introduces a similar amount of code.

Because of this I opted to write a simple home grown solution. After
all, all we need is storing a job ID somewhere so we can efficiently
look it up; we don't need extra web UIs (as provided by sidekiq-status)
or complex APIs to update progress, etc.

This is where Gitlab::SidekiqStatus comes in handy. This namespace
contains some code used for tracking, removing, and looking up job IDs;
all without having to scan over an entire queue. Data is removed
explicitly, but also expires automatically just in case.

Using this API we can now schedule jobs in a fork-join like manner: we
schedule the jobs in Sidekiq, process them in parallel, then wait for
completion. By using Sidekiq we can leverage all the benefits such as
being able to scale across multiple cores and hosts, retrying failed
jobs, etc.

The one downside is that we need to make sure we can deal with
unexpected increases in job processing timings. To deal with this the
class Gitlab::JobWaiter (used for waiting for jobs to complete) will
only wait a number of seconds (30 by default). Once this timeout is
reached it will simply return.

For GitLab.com almost all AuthorizedProjectWorker jobs complete in
seconds, only very rarely do we spike to job timings of around a minute.
These in turn seem to be the result of external factors (e.g. deploys),
in which case a user is most likely not able to use the system anyway.

In short, this new solution should ensure that jobs are processed
properly and that in almost all cases a user has access to their
resources whenever they need to have access.

Gemfile

@@ -322,7 +322,7 @@ group :test do
   gem 'email_spec', '~> 1.6.0'
   gem 'json-schema', '~> 2.6.2'
   gem 'webmock', '~> 1.21.0'
-  gem 'test_after_commit', '~> 0.4.2'
+  gem 'test_after_commit', '~> 1.1'
   gem 'sham_rack', '~> 1.3.6'
   gem 'timecop', '~> 0.8.0'
 end

Gemfile.lock

@@ -760,7 +760,7 @@ GEM
     teaspoon-jasmine (2.2.0)
       teaspoon (>= 1.0.0)
     temple (0.7.7)
-    test_after_commit (0.4.2)
+    test_after_commit (1.1.0)
       activerecord (>= 3.2)
     thin (1.7.0)
       daemons (~> 1.0, >= 1.0.9)
@@ -997,7 +997,7 @@ DEPENDENCIES
   sys-filesystem (~> 1.1.6)
   teaspoon (~> 1.1.0)
   teaspoon-jasmine (~> 2.2.0)
-  test_after_commit (~> 0.4.2)
+  test_after_commit (~> 1.1)
   thin (~> 1.7.0)
   timecop (~> 0.8.0)
   truncato (~> 0.7.8)

app/models/member.rb

@@ -68,9 +68,9 @@ class Member < ActiveRecord::Base
   after_create :send_request, if: :request?, unless: :importing?
   after_create :create_notification_setting, unless: [:pending?, :importing?]
   after_create :post_create_hook, unless: [:pending?, :importing?]
-  after_create :refresh_member_authorized_projects, if: :importing?
   after_update :post_update_hook, unless: [:pending?, :importing?]
   after_destroy :post_destroy_hook, unless: :pending?
+  after_commit :refresh_member_authorized_projects
   delegate :name, :username, :email, to: :user, prefix: true
@@ -147,8 +147,6 @@ class Member < ActiveRecord::Base
         member.save
       end
-      UserProjectAccessChangedService.new(user.id).execute if user.is_a?(User)
-
       member
     end
@@ -275,23 +273,27 @@ class Member < ActiveRecord::Base
   end
   def post_create_hook
-    UserProjectAccessChangedService.new(user.id).execute
     system_hook_service.execute_hooks_for(self, :create)
   end
   def post_update_hook
-    UserProjectAccessChangedService.new(user.id).execute if access_level_changed?
+    # override in sub class
   end
   def post_destroy_hook
-    refresh_member_authorized_projects
     system_hook_service.execute_hooks_for(self, :destroy)
   end
+  # Refreshes authorizations of the current member.
+  #
+  # This method schedules a job using Sidekiq and as such **must not** be called
+  # in a transaction. Doing so can lead to the job running before the
+  # transaction has been committed, resulting in the job either throwing an
+  # error or not doing any meaningful work.
   def refresh_member_authorized_projects
-    # If user/source is being destroyed, project access are gonna be destroyed eventually
-    # because of DB foreign keys, so we shouldn't bother with refreshing after each
-    # member is destroyed through association
+    # If user/source is being destroyed, project access are going to be
+    # destroyed eventually because of DB foreign keys, so we shouldn't bother
+    # with refreshing after each member is destroyed through association
     return if destroyed_by_association.present?
     UserProjectAccessChangedService.new(user_id).execute

app/models/project_group_link.rb

@@ -16,8 +16,7 @@ class ProjectGroupLink < ActiveRecord::Base
   validates :group_access, inclusion: { in: Gitlab::Access.values }, presence: true
   validate :different_group
-  after_create :refresh_group_members_authorized_projects
-  after_destroy :refresh_group_members_authorized_projects
+  after_commit :refresh_group_members_authorized_projects
   def self.access_options
     Gitlab::Access.options

app/services/user_project_access_changed_service.rb

@@ -4,6 +4,6 @@ class UserProjectAccessChangedService
   end
   def execute
-    AuthorizedProjectsWorker.bulk_perform_async(@user_ids.map { |id| [id] })
+    AuthorizedProjectsWorker.bulk_perform_and_wait(@user_ids.map { |id| [id] })
   end
 end

app/workers/authorized_projects_worker.rb

@@ -2,6 +2,13 @@ class AuthorizedProjectsWorker
   include Sidekiq::Worker
   include DedicatedSidekiqQueue
+  # Schedules multiple jobs and waits for them to be completed.
+  def self.bulk_perform_and_wait(args_list)
+    job_ids = bulk_perform_async(args_list)
+
+    Gitlab::JobWaiter.new(job_ids).wait
+  end
+
   def self.bulk_perform_async(args_list)
     Sidekiq::Client.push_bulk('class' => self, 'args' => args_list)
   end

changelogs/unreleased/refresh-authorizations-fork-join.yml

+---
+title: Fix race conditions for AuthorizedProjectsWorker
+merge_request:
+author:

config/initializers/sidekiq.rb

@@ -12,6 +12,11 @@ Sidekiq.configure_server do |config|
     chain.add Gitlab::SidekiqMiddleware::ArgumentsLogger if ENV['SIDEKIQ_LOG_ARGUMENTS']
     chain.add Gitlab::SidekiqMiddleware::MemoryKiller if ENV['SIDEKIQ_MEMORY_KILLER_MAX_RSS']
     chain.add Gitlab::SidekiqMiddleware::RequestStoreMiddleware unless ENV['SIDEKIQ_REQUEST_STORE'] == '0'
+    chain.add Gitlab::SidekiqStatus::ServerMiddleware
+  end
+
+  config.client_middleware do |chain|
+    chain.add Gitlab::SidekiqStatus::ClientMiddleware
   end
   # Sidekiq-cron: load recurring jobs from gitlab.yml
@@ -46,6 +51,10 @@ end
 Sidekiq.configure_client do |config|
   config.redis = redis_config_hash
+
+  config.client_middleware do |chain|
+    chain.add Gitlab::SidekiqStatus::ClientMiddleware
+  end
 end
 # The Sidekiq client API always adds the queue to the Sidekiq queue

db/fixtures/development/01_admin.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   User.seed do |s|
     s.id = 1

db/fixtures/development/04_project.rb

-require 'sidekiq/testing'
+require './spec/support/sidekiq'
 Sidekiq::Testing.inline! do
   Gitlab::Seeder.quiet do

db/fixtures/development/05_users.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   20.times do |i|
     begin

db/fixtures/development/06_teams.rb

-require 'sidekiq/testing'
+require './spec/support/sidekiq'
 Sidekiq::Testing.inline! do
   Gitlab::Seeder.quiet do

db/fixtures/development/07_milestones.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   Project.all.each do |project|
     5.times do |i|

db/fixtures/development/09_issues.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   Project.all.each do |project|
     10.times do

db/fixtures/development/10_merge_requests.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   # Limit the number of merge requests per project to avoid long seeds
   MAX_NUM_MERGE_REQUESTS = 10

db/fixtures/development/11_keys.rb

-Gitlab::Seeder.quiet do
-  User.first(10).each do |user|
-    key = "ssh-rsa AAAAB3NzaC1yc2EAAAABJQAAAIEAiPWx6WM4lhHNedGfBpPJNPpZ7yKu+dnn1SJejgt#{user.id + 100}6k6YjzGGphH2TUxwKzxcKDKKezwkpfnxPkSMkuEspGRt/aZZ9wa++Oi7Qkr8prgHc4soW6NUlfDzpvZK2H5E7eQaSeP3SAwGmQKUFHCddNaP0L+hM7zhFNzjFvpaMgJw0="
-    user.keys.create(
-      title: "Sample key #{user.id}",
-      key: key
-    )
-    print '.'
+require './spec/support/sidekiq'
+# Creating keys runs a gitlab-shell worker. Since we may not have the right
+# gitlab-shell path set (yet) we need to disable this for these fixtures.
+Sidekiq::Testing.disable! do
+  Gitlab::Seeder.quiet do
+    User.first(10).each do |user|
+      key = "ssh-rsa AAAAB3NzaC1yc2EAAAABJQAAAIEAiPWx6WM4lhHNedGfBpPJNPpZ7yKu+dnn1SJejgt#{user.id + 100}6k6YjzGGphH2TUxwKzxcKDKKezwkpfnxPkSMkuEspGRt/aZZ9wa++Oi7Qkr8prgHc4soW6NUlfDzpvZK2H5E7eQaSeP3SAwGmQKUFHCddNaP0L+hM7zhFNzjFvpaMgJw0="
+      user.keys.create(
+        title: "Sample key #{user.id}",
+        key: key
+      )
+
+      print '.'
+    end
   end
 end

db/fixtures/development/12_snippets.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   content =<<eos
 class Member < ActiveRecord::Base

db/fixtures/development/13_comments.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   Issue.all.each do |issue|
     project = issue.project

db/fixtures/development/14_pipelines.rb

+require './spec/support/sidekiq'
+
 class Gitlab::Seeder::Pipelines
   STAGES = %w[build test deploy notify]
   BUILDS = [

db/fixtures/development/15_award_emoji.rb

+require './spec/support/sidekiq'
+
 Gitlab::Seeder.quiet do
   emoji = Gitlab::AwardEmoji.emojis.keys