until_executed lock not releasing after job finishes

[kapso]

I’m seeing persistent Redis lock digests after a job finishes execution using lock: :until_executed in Sidekiq 8.0.4 with sidekiq-unique-jobs 8.0.11 on Rails 8.0.3.

• Job runs and finishes successfully.
• Lock digest (uniquejobs:) remains in Redis after job completion.
• As a result, any subsequent .perform_async calls are rejected
• I’ve verified the job does not raise exceptions.

Worker

# worker
class ProjectWorker
  include Sidekiq::Worker

  sidekiq_options queue: :project
                  retry: 8,
                  lock: :until_executed,
                  on_conflict: :log

  def perform(arg1, arg2)
    puts "Running: #{arg1}, #{arg2}"
  end
end

# enqueue
ProjectWorker.perform_async("hello", "world")

Gem versions:

Sidekiq v8.0.4 / Pro v8.0.1
sidekiq-unique-jobs: 8.0.11
rails: 8.0.3
redis: 7.2.0
ruby: 3.4.4

Sidekiq config:

require 'sidekiq/pro/web'
require 'sidekiq-scheduler'
require 'sidekiq-scheduler/web'
require 'sidekiq-unique-jobs'
require 'middleware/sidekiq/set_current_context'
require 'middleware/sidekiq/account_tenant_scoping'

require 'sidekiq/middleware/current_attributes'
Sidekiq::CurrentAttributes.persist('CurrentPersist')

Sidekiq.configure_client do |config|
  config.redis = { url: Settings.redis.url, ssl_params: { verify_mode: OpenSSL::SSL::VERIFY_NONE } }
  config.logger.level = Rails.env.development? ? Logger::INFO : Logger::WARN

  # Sidekiq Pro options
  Sidekiq::Client.reliable_push! unless Rails.env.test?

  config.client_middleware do |chain|
    chain.add SidekiqUniqueJobs::Middleware::Client
  end
end

Sidekiq.configure_server do |config|
  config.redis = { url: Settings.redis.url, ssl_params: { verify_mode: OpenSSL::SSL::VERIFY_NONE } }
  config.logger.level = Rails.env.development? ? Logger::INFO : Logger::WARN

  # Default is 5
  config.average_scheduled_poll_interval = 2

  config.on(:startup) do
    Sidekiq.schedule = YAML.safe_load(ERB.new(Rails.root.join('config/schedule.yml').read).result)
    SidekiqScheduler::Scheduler.instance.reload_schedule!
  end

  # Sidekiq Pro options
  Sidekiq::Client.reliable_push! unless Rails.env.test?
  config.super_fetch!
  config.reliable_scheduler!

  config.client_middleware do |chain|
    chain.add SidekiqUniqueJobs::Middleware::Client
  end

  config.server_middleware do |chain|
    chain.add SidekiqUniqueJobs::Middleware::Server
    chain.add Middleware::Sidekiq::SetCurrentContext
    chain.add Middleware::Sidekiq::AccountTenantScoping
  end

  SidekiqUniqueJobs::Server.configure(config)
end

unless Rails.env.local?
  Sidekiq::Web.use Rack::Auth::Basic do |username, password|
    username == Settings.sidekiq.admin.username && password == Settings.sidekiq.admin.password
  end
end

# Sidekiq Unique Jobs
SidekiqUniqueJobs.configure do |config|
  config.lock_info = true
end

[kapso]

@mhenrixon curious if this has been identified as a bug or maybe something we are doing wrong? Thanks

[RaulGuichard]

@kapso Did you find a reason for this? I'm having the same problem with no apparent reason, it was working as it was meant to but suddenly locks are not being released, can't figure why.