Squashed commit of the following:

commit 8862cc2b083f59c0a59a2175b047c6b999ac38a7
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Thu Mar 1 18:01:55 2018 +0900

    Change the place of definition of object_storage:archive_legacy_trace

commit 51b26252ff5462c604d852bce9953cd381dcf218
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Thu Mar 1 17:49:37 2018 +0900

    Add object_storage queue to sidekiq_queues.ym. and correct queue name in all_queues.yml.

commit 1313ea164cd59992a469983f8579fe4e2a06e2d8
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Thu Mar 1 17:45:14 2018 +0900

    Fix spec failure on mysql because it can not store long traces

commit d1ba3523139ead0ca7a376c962fd1d7725bf1397
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Thu Mar 1 17:38:44 2018 +0900

    Fix static analysys

commit 9f8c8f9c6bb42c5601c4c15b601256e988871909
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 18:48:08 2018 +0900

    Add spec for workers

commit a34a06fbcc1d14a238b3a0b83e631ac714d8acb2
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 18:43:35 2018 +0900

    Add ObjectStorageQueue concern and test

commit 4203125467f12f32982c454ea50483ad9c04ac62
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 18:43:19 2018 +0900

    Fix rake task to use corrrect SQL

commit 0fb390a4467f868f9c9d9f999589aebeff1e7bec
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 18:14:41 2018 +0900

    Raise an error if conditions are not fulfilled in archive method

commit 87177b4d2c73e63dd4795f3375374dd3ae311c00
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 17:29:24 2018 +0900

    Add changelog

commit 267e1a95fc49f9b926925b3bf03e403f6a5a3dd9
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 03:44:13 2018 +0900

    Add soec for achevie! method. Fixed the method

commit d13208bcd48f701df1e016ea11d7e140b08edafd
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 01:24:23 2018 +0900

    Use find_in_batches for rake task

commit 9362c06bd0fea047be7e050937b4cf3c7caebfdb
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Wed Feb 28 01:24:01 2018 +0900

    Add object_storage queue

commit 8b2d135da2803f89505119a9c2a72b2c50b55396
Author: Shinya Maeda <shinya@gitlab.com>
Date:   Mon Feb 26 23:06:04 2018 +0900

    Add rake task. Adopt the latest fix. Drop CreateTraceArtifactService

commit 8705cd63d1e43039b3e3a2f6d6ecf3fdbe95f0ee
Author: Kamil Trzciński <ayufan@ayufan.eu>
Date:   Fri Feb 23 13:08:38 2018 +0100

    Proper fix for artifacts trace migration which is fully safe

app/services/ci/create_trace_artifact_service.rb

-module Ci
-  class CreateTraceArtifactService < BaseService
-    def execute(job)
-      return if job.job_artifacts_trace
-
-      job.trace.read do |stream|
-        break unless stream.file?
-
-        clone_file!(stream.path, JobArtifactUploader.workhorse_upload_path) do |clone_path|
-          create_job_trace!(job, clone_path)
-          FileUtils.rm(stream.path)
-        end
-      end
-    end
-
-    private
-
-    def create_job_trace!(job, path)
-      File.open(path) do |stream|
-        job.create_job_artifacts_trace!(
-          project: job.project,
-          file_type: :trace,
-          file: stream)
-      end
-    end
-
-    def clone_file!(src_path, temp_dir)
-      FileUtils.mkdir_p(temp_dir)
-      Dir.mktmpdir('tmp-trace', temp_dir) do |dir_path|
-        temp_path = File.join(dir_path, "job.log")
-        FileUtils.copy(src_path, temp_path)
-        yield(temp_path)
-      end
-    end
-  end
-end

app/workers/all_queues.yml

@@ -62,6 +62,8 @@
 - repository_check:repository_check_clear
 - repository_check:repository_check_single_repository
 
+- object_storage:archive_legacy_trace
+
 - default
 - mailers # ActionMailer::DeliveryJob.queue_name
 

app/workers/archive_legacy_trace_worker.rb

+class ArchiveLegacyTraceWorker
+  include ApplicationWorker
+  include ObjectStorageQueue
+
+  def perform(job_id)
+    Ci::Build.find_by(id: job_id).try do |job|
+      job.trace.archive!
+    end
+  end
+end

app/workers/concerns/object_storage_queue.rb

+# Concern for setting Sidekiq settings for the various GitLab ObjectStorage workers.
+module ObjectStorageQueue
+  extend ActiveSupport::Concern
+
+  included do
+    queue_namespace :object_storage
+  end
+end

app/workers/create_trace_artifact_worker.rb

@@ -3,8 +3,8 @@ class CreateTraceArtifactWorker
   include PipelineQueue
 
   def perform(job_id)
-    Ci::Build.preload(:project, :user).find_by(id: job_id).try do |job|
-      Ci::CreateTraceArtifactService.new(job.project, job.user).execute(job)
+    Ci::Build.find_by(id: job_id).try do |job|
+      job.trace.archive!
     end
   end
 end

changelogs/unreleased/proper-fix-for-artifacts-service.yml

+---
+title: Add archive feature to trace
+merge_request: 17314
+author:
+type: added

config/sidekiq_queues.yml

@@ -69,6 +69,7 @@
   - [storage_migrator, 1]
   - [pages_domain_verification, 1]
   - [plugin, 1]
+  - [object_storage, 1]
 
   # EE-specific queues
   - [ldap_group_sync, 2]
@@ -84,4 +85,3 @@
   - [elastic_commit_indexer, 1]
   - [export_csv, 1]
   - [object_storage_upload, 1]
-  - [object_storage, 1]

lib/gitlab/ci/trace.rb

@@ -93,8 +93,52 @@ module Gitlab
         job.erase_old_trace!
       end
 
+      def archive!
+        raise 'Already archived' if trace_artifact
+        raise 'Job is not finished yet' unless job.complete?
+
+        if current_path
+          File.open(current_path) do |stream|
+            archive_stream!(stream)
+            FileUtils.rm(current_path)
+          end
+        elsif old_trace
+          StringIO.new(old_trace, 'rb').tap do |stream|
+            archive_stream!(stream)
+            job.erase_old_trace!
+          end
+        end
+      end
+
       private
 
+      def archive_stream!(stream)
+        clone_file!(stream, JobArtifactUploader.workhorse_upload_path) do |clone_path|
+          create_job_trace!(job, clone_path)
+        end
+      end
+
+      def clone_file!(src_stream, temp_dir)
+        FileUtils.mkdir_p(temp_dir)
+        Dir.mktmpdir('tmp-trace', temp_dir) do |dir_path|
+          temp_path = File.join(dir_path, "job.log")
+          FileUtils.touch(temp_path)
+          size = IO.copy_stream(src_stream, temp_path)
+          raise 'Not all saved' unless size == src_stream.size
+
+          yield(temp_path)
+        end
+      end
+
+      def create_job_trace!(job, path)
+        File.open(path) do |stream|
+          job.create_job_artifacts_trace!(
+            project: job.project,
+            file_type: :trace,
+            file: stream)
+        end
+      end
+
       def ensure_path
         return current_path if current_path
 

lib/tasks/gitlab/traces.rake

+require 'logger'
+require 'resolv-replace'
+
+desc "GitLab | Archive legacy traces to trace artifacts"
+namespace :gitlab do
+  namespace :traces do
+    task archive: :environment do
+      logger = Logger.new(STDOUT)
+      logger.info('Archiving legacy traces')
+
+      Ci::Build.finished
+        .where('NOT EXISTS (?)',
+          Ci::JobArtifact.select(1).trace.where('ci_builds.id = ci_job_artifacts.job_id'))
+        .order(id: :asc)
+        .find_in_batches(batch_size: 1000) do |jobs|
+        job_ids = jobs.map { |job| [job.id] }
+
+        ArchiveLegacyTraceWorker.bulk_perform_async(job_ids)
+
+        logger.info("Scheduled #{job_ids.count} jobs. From #{job_ids.min} to #{job_ids.max}")
+      end
+    end
+  end
+end

spec/lib/gitlab/ci/trace_spec.rb

@@ -399,4 +399,140 @@ describe Gitlab::Ci::Trace do
       end
     end
   end
+
+  describe '#archive!' do
+    subject { trace.archive! }
+
+    shared_examples 'archive trace file' do
+      it do
+        expect { subject }.to change { Ci::JobArtifact.count }.by(1)
+
+        build.reload
+        expect(build.trace.exist?).to be_truthy
+        expect(build.job_artifacts_trace.file.exists?).to be_truthy
+        expect(build.job_artifacts_trace.file.filename).to eq('job.log')
+        expect(File.exist?(src_path)).to be_falsy
+        expect(src_checksum)
+          .to eq(Digest::SHA256.file(build.job_artifacts_trace.file.path).digest)
+      end
+    end
+
+    shared_examples 'source trace file stays intact' do |error:|
+      it do
+        expect { subject }.to raise_error(error)
+
+        build.reload
+        expect(build.trace.exist?).to be_truthy
+        expect(build.job_artifacts_trace).to be_nil
+        expect(File.exist?(src_path)).to be_truthy
+      end
+    end
+
+    shared_examples 'archive trace in database' do
+      it do
+        expect { subject }.to change { Ci::JobArtifact.count }.by(1)
+
+        build.reload
+        expect(build.trace.exist?).to be_truthy
+        expect(build.job_artifacts_trace.file.exists?).to be_truthy
+        expect(build.job_artifacts_trace.file.filename).to eq('job.log')
+        expect(build.old_trace).to be_nil
+        expect(src_checksum)
+          .to eq(Digest::SHA256.file(build.job_artifacts_trace.file.path).digest)
+      end
+    end
+
+    shared_examples 'source trace in database stays intact' do |error:|
+      it do
+        expect { subject }.to raise_error(error)
+
+        build.reload
+        expect(build.trace.exist?).to be_truthy
+        expect(build.job_artifacts_trace).to be_nil
+        expect(build.old_trace).to eq(trace_content)
+      end
+    end
+
+    context 'when job does not have trace artifact' do
+      context 'when trace file stored in default path' do
+        let!(:build) { create(:ci_build, :success, :trace_live) }
+        let!(:src_path) { trace.read { |s| return s.path } }
+        let!(:src_checksum) { Digest::SHA256.file(src_path).digest }
+
+        it_behaves_like 'archive trace file'
+
+        context 'when failed to create clone file' do
+          before do
+            allow_any_instance_of(described_class)
+              .to receive(:clone_file!).and_raise('Not all saved')
+          end
+
+          it_behaves_like 'source trace file stays intact', error: 'Not all saved'
+        end
+
+        context 'when failed to create job artifact record' do
+          before do
+            allow_any_instance_of(Ci::JobArtifact).to receive(:save).and_return(false)
+            allow_any_instance_of(Ci::JobArtifact).to receive_message_chain(:errors, :full_messages)
+              .and_return(%w[Error Error])
+          end
+
+          it_behaves_like 'source trace file stays intact', error: ActiveRecord::RecordInvalid
+        end
+      end
+
+      context 'when trace is stored in database' do
+        let(:build) { create(:ci_build, :success) }
+        let(:trace_content) { 'Sample trace' }
+        let!(:src_checksum) { Digest::SHA256.digest(trace_content) }
+
+        before do
+          build.update_column(:trace, trace_content)
+        end
+
+        it_behaves_like 'archive trace in database'
+
+        context 'when failed to create clone file' do
+          before do
+            allow_any_instance_of(described_class)
+              .to receive(:clone_file!).and_raise('Not all saved')
+          end
+
+          it_behaves_like 'source trace in database stays intact', error: 'Not all saved'
+        end
+
+        context 'when failed to create job artifact record' do
+          before do
+            allow_any_instance_of(Ci::JobArtifact).to receive(:save).and_return(false)
+            allow_any_instance_of(Ci::JobArtifact).to receive_message_chain(:errors, :full_messages)
+              .and_return(%w[Error Error])
+          end
+
+          it_behaves_like 'source trace in database stays intact', error: ActiveRecord::RecordInvalid
+        end
+      end
+    end
+
+    context 'when job has trace artifact' do
+      before do
+        create(:ci_job_artifact, :trace, job: build)
+      end
+
+      it 'does not archive' do
+        expect_any_instance_of(described_class).not_to receive(:archive_stream!)
+        expect { subject }.to raise_error('Already archived')
+        expect(build.job_artifacts_trace.file.exists?).to be_truthy
+      end
+    end
+
+    context 'when job is not finished yet' do
+      let!(:build) { create(:ci_build, :running, :trace_live) }
+
+      it 'does not archive' do
+        expect_any_instance_of(described_class).not_to receive(:archive_stream!)
+        expect { subject }.to raise_error('Job is not finished yet')
+        expect(build.trace.exist?).to be_truthy
+      end
+    end
+  end
 end

spec/services/ci/create_trace_artifact_service_spec.rb

-require 'spec_helper'
-
-describe Ci::CreateTraceArtifactService do
-  describe '#execute' do
-    subject { described_class.new(nil, nil).execute(job) }
-
-    context 'when the job does not have trace artifact' do
-      context 'when the job has a trace file' do
-        let!(:job) { create(:ci_build, :trace_live) }
-        let!(:legacy_path) { job.trace.read { |stream| return stream.path } }
-        let!(:legacy_checksum) { Digest::SHA256.file(legacy_path).hexdigest }
-        let(:new_path) { job.job_artifacts_trace.file.path }
-        let(:new_checksum) { Digest::SHA256.file(new_path).hexdigest }
-
-        it { expect(File.exist?(legacy_path)).to be_truthy }
-
-        it 'creates trace artifact' do
-          expect { subject }.to change { Ci::JobArtifact.count }.by(1)
-
-          expect(File.exist?(legacy_path)).to be_falsy
-          expect(File.exist?(new_path)).to be_truthy
-          expect(new_checksum).to eq(legacy_checksum)
-          expect(job.job_artifacts_trace.file.exists?).to be_truthy
-          expect(job.job_artifacts_trace.file.filename).to eq('job.log')
-        end
-
-        context 'when failed to create trace artifact record' do
-          before do
-            # When ActiveRecord error happens
-            allow_any_instance_of(Ci::JobArtifact).to receive(:save).and_return(false)
-            allow_any_instance_of(Ci::JobArtifact).to receive_message_chain(:errors, :full_messages)
-              .and_return("Error")
-
-            subject rescue nil
-
-            job.reload
-          end
-
-          it 'keeps legacy trace and removes trace artifact' do
-            expect(File.exist?(legacy_path)).to be_truthy
-            expect(job.job_artifacts_trace).to be_nil
-          end
-        end
-      end
-
-      context 'when the job does not have a trace file' do
-        let!(:job) { create(:ci_build) }
-
-        it 'does not create trace artifact' do
-          expect { subject }.not_to change { Ci::JobArtifact.count }
-        end
-      end
-    end
-
-    context 'when the job has already had trace artifact' do
-      let!(:job) { create(:ci_build, :trace_artifact) }
-
-      it 'does not create trace artifact' do
-        expect { subject }.not_to change { Ci::JobArtifact.count }
-      end
-    end
-  end
-end

spec/tasks/gitlab/traces_rake_spec.rb

+require 'rake_helper'
+
+describe 'gitlab:traces rake tasks' do
+  before do
+    Rake.application.rake_require 'tasks/gitlab/traces'
+  end
+
+  shared_examples 'passes the job id to worker' do
+    it do
+      expect(ArchiveLegacyTraceWorker).to receive(:bulk_perform_async).with([[job.id]])
+
+      run_rake_task('gitlab:traces:archive')
+    end
+  end
+
+  shared_examples 'does not pass the job id to worker' do
+    it do
+      expect(ArchiveLegacyTraceWorker).not_to receive(:bulk_perform_async)
+
+      run_rake_task('gitlab:traces:archive')
+    end
+  end
+
+  context 'when trace file stored in default path' do
+    let!(:job) { create(:ci_build, :success, :trace_live) }
+
+    it_behaves_like 'passes the job id to worker'
+  end
+
+  context 'when trace is stored in database' do
+    let!(:job) { create(:ci_build, :success) }
+
+    before do
+      job.update_column(:trace, 'trace in db')
+    end
+
+    it_behaves_like 'passes the job id to worker'
+  end
+
+  context 'when job has trace artifact' do
+    let!(:job) { create(:ci_build, :success) }
+
+    before do
+      create(:ci_job_artifact, :trace, job: job)
+    end
+
+    it_behaves_like 'does not pass the job id to worker'
+  end
+
+  context 'when job is not finished yet' do
+    let!(:build) { create(:ci_build, :running, :trace_live) }
+
+    it_behaves_like 'does not pass the job id to worker'
+  end
+end

spec/workers/archive_legacy_trace_worker_spec.rb

+require 'spec_helper'
+
+describe ArchiveLegacyTraceWorker do
+  it { is_expected.to be_a(ObjectStorageQueue) }
+  it { is_expected.not_to be_a(PipelineQueue) }
+
+  describe '#perform' do
+    subject { described_class.new.perform(job&.id) }
+
+    context 'when job is found' do
+      let(:job) { create(:ci_build) }
+
+      it 'executes service' do
+        expect_any_instance_of(Gitlab::Ci::Trace).to receive(:archive!)
+
+        subject
+      end
+    end
+
+    context 'when job is not found' do
+      let(:job) { nil }
+
+      it 'does not execute service' do
+        expect_any_instance_of(Gitlab::Ci::Trace).not_to receive(:archive!)
+
+        subject
+      end
+    end
+  end
+end

spec/workers/concerns/object_storage_queue_spec.rb

+require 'spec_helper'
+
+describe ObjectStorageQueue do
+  let(:worker) do
+    Class.new do
+      def self.name
+        'DummyWorker'
+      end
+
+      include ApplicationWorker
+      include ObjectStorageQueue
+    end
+  end
+
+  it 'sets a default object storage queue automatically' do
+    expect(worker.sidekiq_options['queue'])
+      .to eq 'object_storage:dummy'
+  end
+end

spec/workers/create_trace_artifact_worker_spec.rb

@@ -8,8 +8,7 @@ describe CreateTraceArtifactWorker do
       let(:job) { create(:ci_build) }
 
       it 'executes service' do
-        expect_any_instance_of(Ci::CreateTraceArtifactService)
-          .to receive(:execute).with(job)
+        expect_any_instance_of(Gitlab::Ci::Trace).to receive(:archive!)
 
         subject
       end
@@ -19,8 +18,7 @@ describe CreateTraceArtifactWorker do
       let(:job) { nil }
 
       it 'does not execute service' do
-        expect_any_instance_of(Ci::CreateTraceArtifactService)
-          .not_to receive(:execute)
+        expect_any_instance_of(Gitlab::Ci::Trace).not_to receive(:archive!)
 
         subject
       end