[ci skip] Add user repository integrity check rake task

doc/raketasks/README.md

@@ -1,10 +1,11 @@
 # Rake tasks
 
 - [Backup restore](backup_restore.md)
+- [Check](check.md)
 - [Cleanup](cleanup.md)
 - [Features](features.md)
 - [Maintenance](maintenance.md) and self-checks
 - [User management](user_management.md)
 - [Web hooks](web_hooks.md)
 - [Import](import.md) of git repositories in bulk
-- [Rebuild authorized_keys file](http://doc.gitlab.com/ce/raketasks/maintenance.html#rebuild-authorized_keys-file) task for administrators
+- [Rebuild authorized_keys file](http://doc.gitlab.com/ce/raketasks/maintenance.html#rebuild-authorized_keys-file) task for administrators

doc/raketasks/check.md

@@ -0,0 +1,63 @@
+# Check Rake Tasks
+
+## Repository Integrity
+
+Even though Git is very resilient and tries to prevent data integrity issues,
+there are times when things go wrong. The following Rake tasks intend to
+help GitLab administrators diagnose problem repositories so they can be fixed.
+
+There are 3 things that are checked to determine integrity.
+
+1. Git repository file system check ([git fsck](https://git-scm.com/docs/git-fsck)).
+   This step verifies the connectivity and validity of objects in the repository.
+1. Check for `config.lock` in the repository directory.
+1. Check for any branch/references lock files in `refs/heads`.
+
+It's important to note that the existence of `config.lock` or reference locks
+alone do not necessarily indicate a problem. Lock files are routinely created
+and removed as Git and GitLab perform operations on the repository. They serve
+to prevent data integrity issues. However, if a Git operation is interrupted these
+locks may not be cleaned up properly.
+
+The following symptoms may indicate a problem with repository integrity. If users
+experience these symptoms you may use the rake tasks described below to determine
+exactly which repositories are causing the trouble.
+
+- Receiving an error when trying to push code - `remote: error: cannot lock ref`
+- A 500 error when viewing the GitLab dashboard or when accessing a specific project.
+
+### Check all GitLab repositories
+
+This task loops through all repositories on the GitLab server and runs the
+3 integrity checks described previously.
+
+```
+# omnibus-gitlab
+sudo gitlab-rake gitlab:repo:check
+
+# installation from source
+bundle exec rake gitlab:repo:check RAILS_ENV=production
+```
+
+### Check repositories for a specific user
+
+This task checks all repositories that a specific user has access to. This is important
+because sometimes you know which user is experiencing trouble but you don't know
+which project might be the cause.
+
+If the rake task is executed without brackets at the end, you will be prompted
+to enter a username.
+
+```bash
+# omnibus-gitlab
+sudo gitlab-rake gitlab:user:check_repos
+sudo gitlab-rake gitlab:user:check_repos[<username>]
+
+# installation from source
+bundle exec rake gitlab:user:check_repos RAILS_ENV=production
+bundle exec rake gitlab:user:check_repos[<username>] RAILS_ENV=production
+```
+
+Example output:
+
+![gitlab:user:check_repos output](check_repos_output.png)

lib/tasks/gitlab/check.rake

@@ -822,10 +822,27 @@ namespace :gitlab do
 
       namespace_dirs.each do |namespace_dir|
         repo_dirs = Dir.glob(File.join(namespace_dir, '*'))
-        repo_dirs.each do |dir|
-          puts "\nChecking repo at #{dir}"
-          system(*%W(#{Gitlab.config.git.bin_path} fsck), chdir: dir)
-        end
+        repo_dirs.each { |repo_dir| check_repo_integrity(repo_dir) }
+      end
+    end
+  end
+
+  namespace :user do
+    desc "GitLab | Check the integrity of a specific user's repositories"
+    task :check_repos, [:username] => :environment do |t, args|
+      username = args[:username] || prompt("Check repository integrity for which username? ".blue)
+      user = User.find_by(username: username)
+      if user
+        repo_dirs = user.authorized_projects.map do |p|
+                      File.join(
+                        Gitlab.config.gitlab_shell.repos_path,
+                        "#{p.path_with_namespace}.git"
+                      )
+                    end
+
+        repo_dirs.each { |repo_dir| check_repo_integrity(repo_dir) }
+      else
+        puts "\nUser '#{username}' not found".red
       end
     end
   end
@@ -952,4 +969,35 @@ namespace :gitlab do
       false
     end
   end
+
+  def check_repo_integrity(repo_dir)
+    puts "\nChecking repo at #{repo_dir.yellow}"
+
+    git_fsck(repo_dir)
+    check_config_lock(repo_dir)
+    check_ref_locks(repo_dir)
+  end
+
+  def git_fsck(repo_dir)
+    puts "Running `git fsck`".yellow
+    system(*%W(#{Gitlab.config.git.bin_path} fsck), chdir: repo_dir)
+  end
+
+  def check_config_lock(repo_dir)
+    config_exists = File.exist?(File.join(repo_dir,'config.lock'))
+    config_output = config_exists ? 'yes'.red : 'no'.green
+    puts "'config.lock' file exists?".yellow + " ... #{config_output}"
+  end
+
+  def check_ref_locks(repo_dir)
+    lock_files = Dir.glob(File.join(repo_dir,'refs/heads/*.lock'))
+    if lock_files.present?
+      puts "Ref lock files exist:".red
+      lock_files.each do |lock_file|
+        puts "  #{lock_file}"
+      end
+    else
+      puts "No ref lock files exist".green
+    end
+  end
 end