DEV: Add ability to log a warning when Sidekiq job runs for too long (#27446)

This commits introduces the `sidekiq_report_long_running_jobs_minutes`
global setting which allows a site administrator to log a warning in the
Rails log when a Sidekiq job has been running for too long.

The warning is logged with the backtrace of the thread that is
processing the Sidekiq job to make it easier to figure out what a
sidekiq job is stuck on.
This commit is contained in:
Alan Guo Xiang Tan 2024-06-13 14:24:44 +08:00 committed by GitHub
parent 8058344f8e
commit ae1d8c50da
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
4 changed files with 169 additions and 0 deletions

View File

@ -190,6 +190,9 @@ serve_static_assets = false
# number of sidekiq workers (launched via unicorn master) # number of sidekiq workers (launched via unicorn master)
sidekiq_workers = 5 sidekiq_workers = 5
# Logs Sidekiq jobs that have been running for longer than the configured number of minutes to the Rails log
sidekiq_report_long_running_jobs_minutes =
# connection reaping helps keep connection counts down, postgres # connection reaping helps keep connection counts down, postgres
# will not work properly with huge numbers of open connections # will not work properly with huge numbers of open connections
# reap connections from pool that are older than 30 seconds # reap connections from pool that are older than 30 seconds

View File

@ -2,6 +2,7 @@
require "sidekiq/pausable" require "sidekiq/pausable"
require "sidekiq_logster_reporter" require "sidekiq_logster_reporter"
require "sidekiq_long_running_job_logger"
Sidekiq.configure_client { |config| config.redis = Discourse.sidekiq_redis_config } Sidekiq.configure_client { |config| config.redis = Discourse.sidekiq_redis_config }
@ -9,6 +10,10 @@ Sidekiq.configure_server do |config|
config.redis = Discourse.sidekiq_redis_config config.redis = Discourse.sidekiq_redis_config
config.server_middleware { |chain| chain.add Sidekiq::Pausable } config.server_middleware { |chain| chain.add Sidekiq::Pausable }
if stuck_sidekiq_job_minutes = GlobalSetting.sidekiq_report_long_running_jobs_minutes
config.on(:startup) { SidekiqLongRunningJobLogger.new(stuck_sidekiq_job_minutes:).start }
end
end end
if Sidekiq.server? if Sidekiq.server?

View File

@ -0,0 +1,81 @@
# frozen_string_literal: true
class SidekiqLongRunningJobLogger
attr_reader :thread
def initialize(stuck_sidekiq_job_minutes:)
@mutex = Mutex.new
@stop_requested = false
# Assume that setting the value of `stuck_sidekiq_job_minutes` lower than 0 is a mistake and set it to 1. This makes
# the code in this class easier to reason about.
@stuck_sidekiq_job_minutes = stuck_sidekiq_job_minutes <= 0 ? 1 : stuck_sidekiq_job_minutes.ceil
end
def start
@thread ||=
begin
hostname = Discourse.os_hostname
seconds_to_sleep_between_checks = (@stuck_sidekiq_job_minutes * 60) / 2
Thread.new do
loop do
break if self.stop_requested?
begin
current_long_running_jobs = Set.new
Sidekiq::Workers.new.each do |process_id, thread_id, work|
next unless process_id.start_with?(hostname)
if Time.at(work["run_at"]).to_i >=
(Time.now - (60 * @stuck_sidekiq_job_minutes)).to_i
next
end
jid = work.dig("payload", "jid")
current_long_running_jobs << jid
next if @seen_long_running_jobs&.include?(jid)
if thread = Thread.list.find { |t| t["sidekiq_tid"] == thread_id }
Rails.logger.warn(<<~MSG)
Sidekiq job `#{work.dig("payload", "class")}` has been running for more than #{@stuck_sidekiq_job_minutes} minutes
#{thread.backtrace.join("\n")}
MSG
end
end
@seen_long_running_jobs = current_long_running_jobs
yield if block_given?
rescue => error
Discourse.warn_exception(
error,
message: "Unexpected error in SidekiqLongRunningJobChecker thread",
)
end
sleep seconds_to_sleep_between_checks
end
end
end
end
# Used for testing to stop the thread. In production, the thread is expected to live for the lifetime of the process.
def stop
@mutex.synchronize { @stop_requested = true }
if @thread
@thread.wakeup
@thread.join
@thread = nil
end
end
private
def stop_requested?
@mutex.synchronize { @stop_requested }
end
end

View File

@ -0,0 +1,80 @@
# frozen_string_literal: true
require "sidekiq_long_running_job_logger"
RSpec.describe SidekiqLongRunningJobLogger do
before do
@orig_logger = Rails.logger
Rails.logger = @fake_logger = FakeLogger.new
end
after { Rails.logger = @orig_logger }
it "logs long-running jobs" do
hostname = Discourse.os_hostname
stuck_sidekiq_job_minutes = 10
Sidekiq::Workers
.expects(:new)
.returns(
[
[
"#{hostname}:1234",
"some_sidekiq_id",
{
"run_at" => (Time.now - (60 * (stuck_sidekiq_job_minutes + 1))).to_i,
"payload" => {
"jid" => "job_1",
"class" => "TestWorker",
},
},
],
[
"#{hostname}:1234",
"some_other_sidekiq_id",
{
"run_at" => Time.now.to_i,
"payload" => {
"jid" => "job_2",
"class" => "AnotherWorker",
},
},
],
],
)
.twice
thread = mock("Thread")
thread.expects(:[]).with("sidekiq_tid").returns("some_sidekiq_id").once
thread.expects(:backtrace).returns(%w[line lines]).once
Thread.expects(:list).returns([thread]).once
begin
checker = described_class.new(stuck_sidekiq_job_minutes:)
loops = 0
checker.start { loops += 1 }
wait_for { loops == 1 }
expect(@fake_logger.warnings.size).to eq(1)
expect(@fake_logger.warnings).to include(
"Sidekiq job `TestWorker` has been running for more than 10 minutes\nline\nlines\n",
)
checker.thread.wakeup # Force the thread to run the next loop
wait_for { loops == 2 }
expect(@fake_logger.warnings.size).to eq(1)
expect(@fake_logger.warnings).to include(
"Sidekiq job `TestWorker` has been running for more than 10 minutes\nline\nlines\n",
)
ensure
checker.stop
end
end
end