Commit 0f602be9 authored by Jacob Vosmaer's avatar Jacob Vosmaer

Clear repository check columns asynchronously

parent 9a30d3b5
...@@ -20,18 +20,14 @@ class Admin::ApplicationSettingsController < Admin::ApplicationController ...@@ -20,18 +20,14 @@ class Admin::ApplicationSettingsController < Admin::ApplicationController
end end
def clear_repository_check_states def clear_repository_check_states
Project.update_all( RepositoryCheck::ClearWorker.perform_async
last_repository_check_failed: false,
last_repository_check_at: nil
)
redirect_to( redirect_to(
admin_application_settings_path, admin_application_settings_path,
notice: 'All repository check states were cleared.' notice: 'Started asynchronous removal of all repository check states.'
) )
end end
private private
def set_application_setting def set_application_setting
......
...@@ -32,7 +32,7 @@ class Admin::ProjectsController < Admin::ApplicationController ...@@ -32,7 +32,7 @@ class Admin::ProjectsController < Admin::ApplicationController
end end
def repository_check def repository_check
SingleRepositoryCheckWorker.perform_async(@project.id) RepositoryCheck::SingleRepositoryWorker.perform_async(@project.id)
redirect_to( redirect_to(
admin_namespace_project_path(@project.namespace, @project), admin_namespace_project_path(@project.namespace, @project),
......
module RepositoryCheck
class BatchWorker
include Sidekiq::Worker
RUN_TIME = 3600
sidekiq_options retry: false
def perform
start = Time.now
# This loop will break after a little more than one hour ('a little
# more' because `git fsck` may take a few minutes), or if it runs out of
# projects to check. By default sidekiq-cron will start a new
# RepositoryCheckWorker each hour so that as long as there are repositories to
# check, only one (or two) will be checked at a time.
project_ids.each do |project_id|
break if Time.now - start >= RUN_TIME
break unless current_settings.repository_checks_enabled
next unless try_obtain_lease(project_id)
SingleRepositoryWorker.new.perform(project_id)
end
end
private
# Project.find_each does not support WHERE clauses and
# Project.find_in_batches does not support ordering. So we just build an
# array of ID's. This is OK because we do it only once an hour, because
# getting ID's from Postgres is not terribly slow, and because no user
# has to sit and wait for this query to finish.
def project_ids
limit = 10_000
never_checked_projects = Project.where('last_repository_check_at IS NULL').limit(limit).
pluck(:id)
old_check_projects = Project.where('last_repository_check_at < ?', 1.week.ago).
reorder('last_repository_check_at ASC').limit(limit).pluck(:id)
never_checked_projects + old_check_projects
end
def try_obtain_lease(id)
# Use a 24-hour timeout because on servers/projects where 'git fsck' is
# super slow we definitely do not want to run it twice in parallel.
Gitlab::ExclusiveLease.new(
"project_repository_check:#{id}",
timeout: 24.hours
).try_obtain
end
def current_settings
# No caching of the settings! If we cache them and an admin disables
# this feature, an active RepositoryCheckWorker would keep going for up
# to 1 hour after the feature was disabled.
if Rails.env.test?
Gitlab::CurrentSettings.fake_application_settings
else
ApplicationSetting.current
end
end
end
end
module RepositoryCheck
class ClearWorker
include Sidekiq::Worker
sidekiq_options retry: false
def perform
# Do batched updates because these updates will be slow and locking
Project.select(:id).find_in_batches(batch_size: 1000) do |batch|
Project.where(id: batch.map(&:id)).update_all(
last_repository_check_failed: nil,
last_repository_check_at: nil,
)
end
end
end
end
\ No newline at end of file
module RepositoryCheck
class SingleRepositoryWorker
include Sidekiq::Worker
sidekiq_options retry: false
def perform(project_id)
project = Project.find(project_id)
project.update_columns(
last_repository_check_failed: !check(project),
last_repository_check_at: Time.now,
)
end
private
def check(project)
# Use 'map do', not 'all? do', to prevent short-circuiting
[project.repository, project.wiki.repository].map do |repository|
git_fsck(repository.path_to_repo)
end.all?
end
def git_fsck(path)
cmd = %W(nice git --git-dir=#{path} fsck)
output, status = Gitlab::Popen.popen(cmd)
if status.zero?
true
else
Gitlab::RepositoryCheckLogger.error("command failed: #{cmd.join(' ')}\n#{output}")
false
end
end
end
end
class RepositoryCheckWorker
include Sidekiq::Worker
RUN_TIME = 3600
sidekiq_options retry: false
def perform
start = Time.now
# This loop will break after a little more than one hour ('a little
# more' because `git fsck` may take a few minutes), or if it runs out of
# projects to check. By default sidekiq-cron will start a new
# RepositoryCheckWorker each hour so that as long as there are repositories to
# check, only one (or two) will be checked at a time.
project_ids.each do |project_id|
break if Time.now - start >= RUN_TIME
break unless current_settings.repository_checks_enabled
next unless try_obtain_lease(project_id)
SingleRepositoryCheckWorker.new.perform(project_id)
end
end
private
# Project.find_each does not support WHERE clauses and
# Project.find_in_batches does not support ordering. So we just build an
# array of ID's. This is OK because we do it only once an hour, because
# getting ID's from Postgres is not terribly slow, and because no user
# has to sit and wait for this query to finish.
def project_ids
limit = 10_000
never_checked_projects = Project.where('last_repository_check_at IS NULL').limit(limit).
pluck(:id)
old_check_projects = Project.where('last_repository_check_at < ?', 1.week.ago).
reorder('last_repository_check_at ASC').limit(limit).pluck(:id)
never_checked_projects + old_check_projects
end
def try_obtain_lease(id)
# Use a 24-hour timeout because on servers/projects where 'git fsck' is
# super slow we definitely do not want to run it twice in parallel.
Gitlab::ExclusiveLease.new(
"project_repository_check:#{id}",
timeout: 24.hours
).try_obtain
end
def current_settings
# No caching of the settings! If we cache them and an admin disables
# this feature, an active RepositoryCheckWorker would keep going for up
# to 1 hour after the feature was disabled.
if Rails.env.test?
Gitlab::CurrentSettings.fake_application_settings
else
ApplicationSetting.current
end
end
end
class SingleRepositoryCheckWorker
include Sidekiq::Worker
sidekiq_options retry: false
def perform(project_id)
project = Project.find(project_id)
project.update_columns(
last_repository_check_failed: !check(project),
last_repository_check_at: Time.now,
)
end
private
def check(project)
# Use 'map do', not 'all? do', to prevent short-circuiting
[project.repository, project.wiki.repository].map do |repository|
git_fsck(repository.path_to_repo)
end.all?
end
def git_fsck(path)
cmd = %W(nice git --git-dir=#{path} fsck)
output, status = Gitlab::Popen.popen(cmd)
if status.zero?
true
else
Gitlab::RepositoryCheckLogger.error("command failed: #{cmd.join(' ')}\n#{output}")
false
end
end
end
...@@ -241,7 +241,7 @@ Settings.cron_jobs['stuck_ci_builds_worker']['cron'] ||= '0 0 * * *' ...@@ -241,7 +241,7 @@ Settings.cron_jobs['stuck_ci_builds_worker']['cron'] ||= '0 0 * * *'
Settings.cron_jobs['stuck_ci_builds_worker']['job_class'] = 'StuckCiBuildsWorker' Settings.cron_jobs['stuck_ci_builds_worker']['job_class'] = 'StuckCiBuildsWorker'
Settings.cron_jobs['repository_check_worker'] ||= Settingslogic.new({}) Settings.cron_jobs['repository_check_worker'] ||= Settingslogic.new({})
Settings.cron_jobs['repository_check_worker']['cron'] ||= '20 * * * *' Settings.cron_jobs['repository_check_worker']['cron'] ||= '20 * * * *'
Settings.cron_jobs['repository_check_worker']['job_class'] = 'RepositoryCheckWorker' Settings.cron_jobs['repository_check_worker']['job_class'] = 'RepositoryCheck::BatchWorker'
Settings.cron_jobs['admin_email_worker'] ||= Settingslogic.new({}) Settings.cron_jobs['admin_email_worker'] ||= Settingslogic.new({})
Settings.cron_jobs['admin_email_worker']['cron'] ||= '0 0 * * *' Settings.cron_jobs['admin_email_worker']['cron'] ||= '0 0 * * *'
Settings.cron_jobs['admin_email_worker']['job_class'] = 'AdminEmailWorker' Settings.cron_jobs['admin_email_worker']['job_class'] = 'AdminEmailWorker'
......
require 'rails_helper' require 'rails_helper'
feature 'Admin uses repository checks', feature: true do feature 'Admin uses repository checks', feature: true do
before do before { login_as :admin }
login_as :admin
end
scenario 'to trigger a single check' do scenario 'to trigger a single check' do
project = create(:empty_project) project = create(:empty_project)
...@@ -17,7 +15,12 @@ feature 'Admin uses repository checks', feature: true do ...@@ -17,7 +15,12 @@ feature 'Admin uses repository checks', feature: true do
end end
scenario 'to see a single failed repository check' do scenario 'to see a single failed repository check' do
visit_admin_project_page(broken_project) project = create(:empty_project)
project.update_columns(
last_repository_check_failed: true,
last_repository_check_at: Time.now,
)
visit_admin_project_page(project)
page.within('.alert') do page.within('.alert') do
expect(page.text).to match(/Last repository check \(.* ago\) failed/) expect(page.text).to match(/Last repository check \(.* ago\) failed/)
...@@ -25,24 +28,16 @@ feature 'Admin uses repository checks', feature: true do ...@@ -25,24 +28,16 @@ feature 'Admin uses repository checks', feature: true do
end end
scenario 'to clear all repository checks', js: true do scenario 'to clear all repository checks', js: true do
project = broken_project
visit admin_application_settings_path visit admin_application_settings_path
click_link 'Clear all repository checks' # pop-up should be auto confirmed expect(RepositoryCheck::ClearWorker).to receive(:perform_async)
click_link 'Clear all repository checks'
expect(project.reload.last_repository_check_failed).to eq(false) expect(page).to have_content('Started asynchronous removal of all repository check states.')
end end
def visit_admin_project_page(project) def visit_admin_project_page(project)
visit admin_namespace_project_path(project.namespace, project) visit admin_namespace_project_path(project.namespace, project)
end end
def broken_project
project = create(:empty_project)
project.update_columns(
last_repository_check_failed: true,
last_repository_check_at: Time.now,
)
project
end
end end
require 'spec_helper' require 'spec_helper'
describe RepositoryCheckWorker do describe RepositoryCheck::BatchWorker do
subject { RepositoryCheckWorker.new } subject { described_class.new }
it 'prefers projects that have never been checked' do it 'prefers projects that have never been checked' do
projects = create_list(:project, 3) projects = create_list(:project, 3)
......
require 'spec_helper'
describe RepositoryCheck::ClearWorker do
it 'clears repository check columns' do
project = create(:empty_project)
project.update_columns(
last_repository_check_failed: true,
last_repository_check_at: Time.now,
)
described_class.new.perform
project.reload
expect(project.last_repository_check_failed).to be_nil
expect(project.last_repository_check_at).to be_nil
end
end
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment