Commit c9077a0e authored by Shinya Maeda's avatar Shinya Maeda Committed by Alessio Caiazza

Add cleanup mechanizm for stale scheduled jobs

parent f76f6df1
...@@ -49,7 +49,8 @@ class CommitStatus < ActiveRecord::Base ...@@ -49,7 +49,8 @@ class CommitStatus < ActiveRecord::Base
stuck_or_timeout_failure: 3, stuck_or_timeout_failure: 3,
runner_system_failure: 4, runner_system_failure: 4,
missing_dependency_failure: 5, missing_dependency_failure: 5,
runner_unsupported: 6 runner_unsupported: 6,
schedule_expired: 7
} }
## ##
......
...@@ -8,6 +8,7 @@ class StuckCiJobsWorker ...@@ -8,6 +8,7 @@ class StuckCiJobsWorker
BUILD_RUNNING_OUTDATED_TIMEOUT = 1.hour BUILD_RUNNING_OUTDATED_TIMEOUT = 1.hour
BUILD_PENDING_OUTDATED_TIMEOUT = 1.day BUILD_PENDING_OUTDATED_TIMEOUT = 1.day
BUILD_SCHEDULED_OUTDATED_TIMEOUT = 1.hour
BUILD_PENDING_STUCK_TIMEOUT = 1.hour BUILD_PENDING_STUCK_TIMEOUT = 1.hour
def perform def perform
...@@ -15,9 +16,10 @@ class StuckCiJobsWorker ...@@ -15,9 +16,10 @@ class StuckCiJobsWorker
Rails.logger.info "#{self.class}: Cleaning stuck builds" Rails.logger.info "#{self.class}: Cleaning stuck builds"
drop :running, BUILD_RUNNING_OUTDATED_TIMEOUT drop :running, :updated_at, BUILD_RUNNING_OUTDATED_TIMEOUT, :stuck_or_timeout_failure
drop :pending, BUILD_PENDING_OUTDATED_TIMEOUT drop :pending, :updated_at, BUILD_PENDING_OUTDATED_TIMEOUT, :stuck_or_timeout_failure
drop_stuck :pending, BUILD_PENDING_STUCK_TIMEOUT drop :scheduled, :scheduled_at, BUILD_SCHEDULED_OUTDATED_TIMEOUT, :schedule_expired
drop_stuck :pending, :updated_at, BUILD_PENDING_STUCK_TIMEOUT, :stuck_or_timeout_failure
remove_lease remove_lease
end end
...@@ -32,25 +34,27 @@ class StuckCiJobsWorker ...@@ -32,25 +34,27 @@ class StuckCiJobsWorker
Gitlab::ExclusiveLease.cancel(EXCLUSIVE_LEASE_KEY, @uuid) Gitlab::ExclusiveLease.cancel(EXCLUSIVE_LEASE_KEY, @uuid)
end end
def drop(status, timeout) def drop(status, column, timeout, reason)
search(status, timeout) do |build| search(status, column, timeout) do |build|
drop_build :outdated, build, status, timeout drop_build :outdated, build, status, timeout, reason
end end
end end
def drop_stuck(status, timeout) def drop_stuck(status, column, timeout, reason)
search(status, timeout) do |build| search(status, column, timeout) do |build|
break unless build.stuck? break unless build.stuck?
drop_build :stuck, build, status, timeout drop_build :stuck, build, status, timeout, reason
end end
end end
# rubocop: disable CodeReuse/ActiveRecord # rubocop: disable CodeReuse/ActiveRecord
def search(status, timeout) def search(status, column, timeout)
quoted_column = ActiveRecord::Base.connection.quote_column_name(column)
loop do loop do
jobs = Ci::Build.where(status: status) jobs = Ci::Build.where(status: status)
.where('ci_builds.updated_at < ?', timeout.ago) .where("#{quoted_column} < ?", timeout.ago)
.includes(:tags, :runner, project: :namespace) .includes(:tags, :runner, project: :namespace)
.limit(100) .limit(100)
.to_a .to_a
...@@ -63,10 +67,10 @@ class StuckCiJobsWorker ...@@ -63,10 +67,10 @@ class StuckCiJobsWorker
end end
# rubocop: enable CodeReuse/ActiveRecord # rubocop: enable CodeReuse/ActiveRecord
def drop_build(type, build, status, timeout) def drop_build(type, build, status, timeout, reason)
Rails.logger.info "#{self.class}: Dropping #{type} build #{build.id} for runner #{build.runner_id} (status: #{status}, timeout: #{timeout})" Rails.logger.info "#{self.class}: Dropping #{type} build #{build.id} for runner #{build.runner_id} (status: #{status}, timeout: #{timeout})"
Gitlab::OptimisticLocking.retry_lock(build, 3) do |b| Gitlab::OptimisticLocking.retry_lock(build, 3) do |b|
b.drop(:stuck_or_timeout_failure) b.drop(reason)
end end
end end
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