diff options
author | GitLab Bot <gitlab-bot@gitlab.com> | 2020-11-19 08:27:35 +0000 |
---|---|---|
committer | GitLab Bot <gitlab-bot@gitlab.com> | 2020-11-19 08:27:35 +0000 |
commit | 7e9c479f7de77702622631cff2628a9c8dcbc627 (patch) | |
tree | c8f718a08e110ad7e1894510980d2155a6549197 /lib/gitlab/background_migration | |
parent | e852b0ae16db4052c1c567d9efa4facc81146e88 (diff) | |
download | gitlab-ce-7e9c479f7de77702622631cff2628a9c8dcbc627.tar.gz |
Add latest changes from gitlab-org/gitlab@13-6-stable-eev13.6.0-rc42
Diffstat (limited to 'lib/gitlab/background_migration')
7 files changed, 429 insertions, 7 deletions
diff --git a/lib/gitlab/background_migration/backfill_design_internal_ids.rb b/lib/gitlab/background_migration/backfill_design_internal_ids.rb new file mode 100644 index 00000000000..553571d5d00 --- /dev/null +++ b/lib/gitlab/background_migration/backfill_design_internal_ids.rb @@ -0,0 +1,130 @@ +# frozen_string_literal: true + +module Gitlab + module BackgroundMigration + # Backfill design.iid for a range of projects + class BackfillDesignInternalIds + # See app/models/internal_id + # This is a direct copy of the application code with the following changes: + # - usage enum is hard-coded to the value for design_management_designs + # - init is not passed around, but ignored + class InternalId < ActiveRecord::Base + def self.track_greatest(subject, scope, new_value) + InternalIdGenerator.new(subject, scope).track_greatest(new_value) + end + + # Increments #last_value with new_value if it is greater than the current, + # and saves the record + # + # The operation locks the record and gathers a `ROW SHARE` lock (in PostgreSQL). + # As such, the increment is atomic and safe to be called concurrently. + def track_greatest_and_save!(new_value) + update_and_save { self.last_value = [last_value || 0, new_value].max } + end + + private + + def update_and_save(&block) + lock! + yield + # update_and_save_counter.increment(usage: usage, changed: last_value_changed?) + save! + last_value + end + end + + # See app/models/internal_id + class InternalIdGenerator + attr_reader :subject, :scope, :scope_attrs + + def initialize(subject, scope) + @subject = subject + @scope = scope + + raise ArgumentError, 'Scope is not well-defined, need at least one column for scope (given: 0)' if scope.empty? + end + + # Create a record in internal_ids if one does not yet exist + # and set its new_value if it is higher than the current last_value + # + # Note this will acquire a ROW SHARE lock on the InternalId record + def track_greatest(new_value) + subject.transaction do + record.track_greatest_and_save!(new_value) + end + end + + def record + @record ||= (lookup || create_record) + end + + def lookup + InternalId.find_by(**scope, usage: usage_value) + end + + def usage_value + 10 # see Enums::InternalId - this is the value for design_management_designs + end + + # Create InternalId record for (scope, usage) combination, if it doesn't exist + # + # We blindly insert without synchronization. If another process + # was faster in doing this, we'll realize once we hit the unique key constraint + # violation. We can safely roll-back the nested transaction and perform + # a lookup instead to retrieve the record. + def create_record + subject.transaction(requires_new: true) do + InternalId.create!( + **scope, + usage: usage_value, + last_value: 0 + ) + end + rescue ActiveRecord::RecordNotUnique + lookup + end + end + + attr_reader :design_class + + def initialize(design_class) + @design_class = design_class + end + + def perform(relation) + start_id, end_id = relation.pluck("min(project_id), max(project_id)").flatten + table = 'design_management_designs' + + ActiveRecord::Base.connection.execute <<~SQL + WITH + starting_iids(project_id, iid) as ( + SELECT project_id, MAX(COALESCE(iid, 0)) + FROM #{table} + WHERE project_id BETWEEN #{start_id} AND #{end_id} + GROUP BY project_id + ), + with_calculated_iid(id, iid) as ( + SELECT design.id, + init.iid + ROW_NUMBER() OVER (PARTITION BY design.project_id ORDER BY design.id ASC) + FROM #{table} as design, starting_iids as init + WHERE design.project_id BETWEEN #{start_id} AND #{end_id} + AND design.iid IS NULL + AND init.project_id = design.project_id + ) + + UPDATE #{table} + SET iid = with_calculated_iid.iid + FROM with_calculated_iid + WHERE #{table}.id = with_calculated_iid.id + SQL + + # track the new greatest IID value + relation.each do |design| + current_max = design_class.where(project_id: design.project_id).maximum(:iid) + scope = { project_id: design.project_id } + InternalId.track_greatest(design, scope, current_max) + end + end + end + end +end diff --git a/lib/gitlab/background_migration/backfill_jira_tracker_deployment_type2.rb b/lib/gitlab/background_migration/backfill_jira_tracker_deployment_type2.rb new file mode 100644 index 00000000000..61145f6a445 --- /dev/null +++ b/lib/gitlab/background_migration/backfill_jira_tracker_deployment_type2.rb @@ -0,0 +1,86 @@ +# frozen_string_literal: true + +# Based on https://community.developer.atlassian.com/t/get-rest-api-3-filter-search/29459/2, +# it's enough at the moment to simply notice if the url is from `atlassian.net` +module Gitlab + module BackgroundMigration + # Backfill the deployment_type in jira_tracker_data table + class BackfillJiraTrackerDeploymentType2 + # Migration only version of jira_tracker_data table + class JiraTrackerDataTemp < ApplicationRecord + self.table_name = 'jira_tracker_data' + + def self.encryption_options + { + key: Settings.attr_encrypted_db_key_base_32, + encode: true, + mode: :per_attribute_iv, + algorithm: 'aes-256-gcm' + } + end + + attr_encrypted :url, encryption_options + attr_encrypted :api_url, encryption_options + + enum deployment_type: { unknown: 0, server: 1, cloud: 2 }, _prefix: :deployment + end + + # Migration only version of services table + class JiraServiceTemp < ApplicationRecord + self.table_name = 'services' + self.inheritance_column = :_type_disabled + end + + def perform(start_id, stop_id) + @server_ids = [] + @cloud_ids = [] + + JiraTrackerDataTemp + .where(id: start_id..stop_id, deployment_type: 0) + .each do |jira_tracker_data| + collect_deployment_type(jira_tracker_data) + end + + unless cloud_ids.empty? + JiraTrackerDataTemp.where(id: cloud_ids) + .update_all(deployment_type: JiraTrackerDataTemp.deployment_types[:cloud]) + end + + unless server_ids.empty? + JiraTrackerDataTemp.where(id: server_ids) + .update_all(deployment_type: JiraTrackerDataTemp.deployment_types[:server]) + end + + mark_jobs_as_succeeded(start_id, stop_id) + end + + private + + attr_reader :server_ids, :cloud_ids + + def client_url(jira_tracker_data) + jira_tracker_data.api_url.presence || jira_tracker_data.url.presence + end + + def server_type(url) + url.downcase.include?('.atlassian.net') ? :cloud : :server + end + + def collect_deployment_type(jira_tracker_data) + url = client_url(jira_tracker_data) + return unless url + + case server_type(url) + when :cloud + cloud_ids << jira_tracker_data.id + else + server_ids << jira_tracker_data.id + end + end + + def mark_jobs_as_succeeded(*arguments) + Gitlab::Database::BackgroundMigrationJob.mark_all_as_succeeded(self.class.name, arguments) + end + end + end +end diff --git a/lib/gitlab/background_migration/backfill_merge_request_cleanup_schedules.rb b/lib/gitlab/background_migration/backfill_merge_request_cleanup_schedules.rb new file mode 100644 index 00000000000..8a58cf9b302 --- /dev/null +++ b/lib/gitlab/background_migration/backfill_merge_request_cleanup_schedules.rb @@ -0,0 +1,40 @@ +# frozen_string_literal: true + +module Gitlab + module BackgroundMigration + # Backfill merge request cleanup schedules of closed/merged merge requests + # without any corresponding records. + class BackfillMergeRequestCleanupSchedules + # Model used for migration added in https://gitlab.com/gitlab-org/gitlab/-/merge_requests/46782. + class MergeRequest < ActiveRecord::Base + include EachBatch + + self.table_name = 'merge_requests' + + def self.eligible + where('merge_requests.state_id IN (2, 3)') + end + end + + def perform(start_id, end_id) + eligible_mrs = MergeRequest.eligible.where(id: start_id..end_id) + scheduled_at_column = "COALESCE(metrics.merged_at, COALESCE(metrics.latest_closed_at, merge_requests.updated_at)) + interval '14 days'" + query = + eligible_mrs + .select("merge_requests.id, #{scheduled_at_column}, NOW(), NOW()") + .joins('LEFT JOIN merge_request_metrics metrics ON metrics.merge_request_id = merge_requests.id') + + result = ActiveRecord::Base.connection.execute <<~SQL + INSERT INTO merge_request_cleanup_schedules (merge_request_id, scheduled_at, created_at, updated_at) + #{query.to_sql} + ON CONFLICT (merge_request_id) DO NOTHING; + SQL + + ::Gitlab::BackgroundMigration::Logger.info( + message: 'Backfilled merge_request_cleanup_schedules records', + count: result.cmd_tuples + ) + end + end + end +end diff --git a/lib/gitlab/background_migration/populate_has_vulnerabilities.rb b/lib/gitlab/background_migration/populate_has_vulnerabilities.rb new file mode 100644 index 00000000000..78140b768fc --- /dev/null +++ b/lib/gitlab/background_migration/populate_has_vulnerabilities.rb @@ -0,0 +1,62 @@ +# frozen_string_literal: true + +module Gitlab + module BackgroundMigration + # This class populates missing dismissal information for + # vulnerability entries. + class PopulateHasVulnerabilities + class ProjectSetting < ActiveRecord::Base # rubocop:disable Style/Documentation + self.table_name = 'project_settings' + + UPSERT_SQL = <<~SQL + WITH upsert_data (project_id, has_vulnerabilities, created_at, updated_at) AS ( + SELECT projects.id, true, current_timestamp, current_timestamp FROM projects WHERE projects.id IN (%{project_ids}) + ) + INSERT INTO project_settings + (project_id, has_vulnerabilities, created_at, updated_at) + (SELECT * FROM upsert_data) + ON CONFLICT (project_id) + DO UPDATE SET + has_vulnerabilities = true, + updated_at = EXCLUDED.updated_at + SQL + + def self.upsert_for(project_ids) + connection.execute(UPSERT_SQL % { project_ids: project_ids.join(', ') }) + end + end + + class Vulnerability < ActiveRecord::Base # rubocop:disable Style/Documentation + include EachBatch + + self.table_name = 'vulnerabilities' + end + + def perform(*project_ids) + ProjectSetting.upsert_for(project_ids) + rescue StandardError => e + log_error(e, project_ids) + ensure + log_info(project_ids) + end + + private + + def log_error(error, project_ids) + ::Gitlab::BackgroundMigration::Logger.error( + migrator: self.class.name, + message: error.message, + project_ids: project_ids + ) + end + + def log_info(project_ids) + ::Gitlab::BackgroundMigration::Logger.info( + migrator: self.class.name, + message: 'Projects has been processed to populate `has_vulnerabilities` information', + count: project_ids.length + ) + end + end + end +end diff --git a/lib/gitlab/background_migration/populate_missing_vulnerability_dismissal_information.rb b/lib/gitlab/background_migration/populate_missing_vulnerability_dismissal_information.rb new file mode 100644 index 00000000000..bc0a181a06c --- /dev/null +++ b/lib/gitlab/background_migration/populate_missing_vulnerability_dismissal_information.rb @@ -0,0 +1,86 @@ +# frozen_string_literal: true + +module Gitlab + module BackgroundMigration + # This class populates missing dismissal information for + # vulnerability entries. + class PopulateMissingVulnerabilityDismissalInformation + class Vulnerability < ActiveRecord::Base # rubocop:disable Style/Documentation + include EachBatch + + self.table_name = 'vulnerabilities' + + has_one :finding, class_name: '::Gitlab::BackgroundMigration::PopulateMissingVulnerabilityDismissalInformation::Finding' + + scope :broken, -> { where('state = 2 AND (dismissed_at IS NULL OR dismissed_by_id IS NULL)') } + + def copy_dismissal_information + return unless finding&.dismissal_feedback + + update_columns( + dismissed_at: finding.dismissal_feedback.created_at, + dismissed_by_id: finding.dismissal_feedback.author_id + ) + end + end + + class Finding < ActiveRecord::Base # rubocop:disable Style/Documentation + include ShaAttribute + + self.table_name = 'vulnerability_occurrences' + + sha_attribute :project_fingerprint + + def dismissal_feedback + Feedback.dismissal.where(category: report_type, project_fingerprint: project_fingerprint, project_id: project_id).first + end + end + + class Feedback < ActiveRecord::Base # rubocop:disable Style/Documentation + DISMISSAL_TYPE = 0 + + self.table_name = 'vulnerability_feedback' + + scope :dismissal, -> { where(feedback_type: DISMISSAL_TYPE) } + end + + def perform(*vulnerability_ids) + Vulnerability.includes(:finding).where(id: vulnerability_ids).each { |vulnerability| populate_for(vulnerability) } + + log_info(vulnerability_ids) + end + + private + + def populate_for(vulnerability) + log_warning(vulnerability) unless vulnerability.copy_dismissal_information + rescue StandardError => error + log_error(error, vulnerability) + end + + def log_info(vulnerability_ids) + ::Gitlab::BackgroundMigration::Logger.info( + migrator: self.class.name, + message: 'Dismissal information has been copied', + count: vulnerability_ids.length + ) + end + + def log_warning(vulnerability) + ::Gitlab::BackgroundMigration::Logger.warn( + migrator: self.class.name, + message: 'Could not update vulnerability!', + vulnerability_id: vulnerability.id + ) + end + + def log_error(error, vulnerability) + ::Gitlab::BackgroundMigration::Logger.error( + migrator: self.class.name, + message: error.message, + vulnerability_id: vulnerability.id + ) + end + end + end +end diff --git a/lib/gitlab/background_migration/populate_vulnerability_feedback_pipeline_id.rb b/lib/gitlab/background_migration/populate_vulnerability_feedback_pipeline_id.rb new file mode 100644 index 00000000000..fc79f7125e3 --- /dev/null +++ b/lib/gitlab/background_migration/populate_vulnerability_feedback_pipeline_id.rb @@ -0,0 +1,13 @@ +# frozen_string_literal: true + +module Gitlab + module BackgroundMigration + # This class updates vulnerability feedback entities with no pipeline id assigned. + class PopulateVulnerabilityFeedbackPipelineId + def perform(project_ids) + end + end + end +end + +Gitlab::BackgroundMigration::PopulateVulnerabilityFeedbackPipelineId.prepend_if_ee('EE::Gitlab::BackgroundMigration::PopulateVulnerabilityFeedbackPipelineId') diff --git a/lib/gitlab/background_migration/replace_blocked_by_links.rb b/lib/gitlab/background_migration/replace_blocked_by_links.rb index 26626aaef79..0c29887bb00 100644 --- a/lib/gitlab/background_migration/replace_blocked_by_links.rb +++ b/lib/gitlab/background_migration/replace_blocked_by_links.rb @@ -12,14 +12,19 @@ module Gitlab blocked_by_links = IssueLink.where(id: start_id..stop_id).where(link_type: 2) ActiveRecord::Base.transaction do - # if there is duplicit bi-directional relation (issue2 is blocked by issue1 - # and issue1 already links issue2), then we can just delete 'blocked by'. - # This should be rare as we have a pre-create check which checks if issues are - # already linked - blocked_by_links + # There could be two edge cases: + # 1) issue1 is blocked by issue2 AND issue2 blocks issue1 (type 1) + # 2) issue1 is blocked by issue2 AND issue2 is related to issue1 (type 0) + # In both cases cases we couldn't convert blocked by relation to + # `issue2 blocks issue` because there is already a link with the same + # source/target id. To avoid these conflicts, we first delete any + # "opposite" links before we update `blocked by` relation. This + # should be rare as we have a pre-create check which checks if issues + # are already linked + opposite_ids = blocked_by_links + .select('opposite_links.id') .joins('INNER JOIN issue_links as opposite_links ON issue_links.source_id = opposite_links.target_id AND issue_links.target_id = opposite_links.source_id') - .where('opposite_links.link_type': 1) - .delete_all + IssueLink.where(id: opposite_ids).delete_all blocked_by_links.update_all('source_id=target_id,target_id=source_id,link_type=1') end |