From 7c6beae91ec037ccec14b5136768ea08d034d165 Mon Sep 17 00:00:00 2001 From: Carmine Paolino Date: Wed, 22 Jul 2026 20:32:26 +0200 Subject: [PATCH 1/4] Add fiber execution mode for workers Workers can now be configured with `fibers: N` as an alternative to `threads: N` (specifying both is invalid). In fiber mode, claimed jobs execute as fibers on a single Async reactor thread, bounded by the configured fiber count. This suits mostly I/O-bound jobs, where cooperative scheduling allows much higher in-flight capacity with fewer database connections than a thread pool. The existing pool was extracted into ExecutionPools::ThreadPool (SolidQueue::Pool remains as an alias), with the new ExecutionPools::FiberPool sharing the same interface. Pools are built on worker boot, after forking, since a reactor thread wouldn't survive a fork. Fiber mode requires the async gem (>= 2.24), which is not a runtime dependency: workers configured with `fibers` raise a clear error if it isn't available. It also requires fiber-scoped isolated execution state (ActiveSupport::IsolatedExecutionState.isolation_level = :fiber), validated at boot and configuration-check time. The database pool size guidance is fiber-aware and Rails-version aware: on Rails 7.2+, fibers release connections between queries, so a fiber worker needs far fewer connections than its fiber count. --- Gemfile.lock | 18 ++ README.md | 34 ++- lib/solid_queue/configuration.rb | 80 +++++- lib/solid_queue/execution_pools.rb | 18 ++ lib/solid_queue/execution_pools/fiber_pool.rb | 239 ++++++++++++++++++ .../execution_pools/thread_pool.rb | 82 ++++++ lib/solid_queue/pool.rb | 51 +--- lib/solid_queue/worker.rb | 67 ++++- solid_queue.gemspec | 1 + .../async_processes_lifecycle_test.rb | 15 +- test/integration/instrumentation_test.rb | 2 +- test/test_helper.rb | 13 + test/unit/async_supervisor_test.rb | 2 +- test/unit/configuration_test.rb | 85 ++++++- test/unit/execution_pools/fiber_pool_test.rb | 135 ++++++++++ test/unit/worker_test.rb | 194 +++++++++++--- 16 files changed, 921 insertions(+), 115 deletions(-) create mode 100644 lib/solid_queue/execution_pools.rb create mode 100644 lib/solid_queue/execution_pools/fiber_pool.rb create mode 100644 lib/solid_queue/execution_pools/thread_pool.rb create mode 100644 test/unit/execution_pools/fiber_pool_test.rb diff --git a/Gemfile.lock b/Gemfile.lock index 7c4662de5..f1cda3c92 100644 --- a/Gemfile.lock +++ b/Gemfile.lock @@ -55,12 +55,22 @@ GEM rake thor (>= 0.14.0) ast (2.4.2) + async (2.38.1) + console (~> 1.29) + fiber-annotation + io-event (~> 1.11) + metrics (~> 0.12) + traces (~> 0.18) base64 (0.3.0) benchmark (0.4.1) bigdecimal (3.3.1) builder (3.3.0) concurrent-ruby (1.3.5) connection_pool (2.5.4) + console (1.34.3) + fiber-annotation + fiber-local (~> 1.1) + json crass (1.0.6) date (3.4.1) debug (1.9.2) @@ -70,6 +80,10 @@ GEM erubi (1.13.1) et-orbi (1.2.11) tzinfo + fiber-annotation (0.2.0) + fiber-local (1.1.0) + fiber-storage + fiber-storage (1.0.1) fugit (1.11.1) et-orbi (~> 1, >= 1.2.11) raabro (~> 1.4) @@ -78,6 +92,7 @@ GEM i18n (1.14.7) concurrent-ruby (~> 1.0) io-console (0.8.0) + io-event (1.14.5) irb (1.14.3) rdoc (>= 4.0.0) reline (>= 0.4.2) @@ -87,6 +102,7 @@ GEM loofah (2.23.1) crass (~> 1.0.2) nokogiri (>= 1.12.0) + metrics (0.15.0) minitest (5.26.0) mocha (2.1.0) ruby2_keywords (>= 0.0.5) @@ -181,6 +197,7 @@ GEM stringio (3.1.2) thor (1.3.2) timeout (0.4.3) + traces (0.18.2) tzinfo (2.0.6) concurrent-ruby (~> 1.0) unicode-display_width (3.1.3) @@ -199,6 +216,7 @@ PLATFORMS DEPENDENCIES appraisal + async (>= 2.24) debug (~> 1.9) logger minitest (~> 5.0) diff --git a/README.md b/README.md index 02614847a..005b5ea5c 100644 --- a/README.md +++ b/README.md @@ -204,6 +204,10 @@ Or you can also set the environment variable `SOLID_QUEUE_SUPERVISOR_MODE` to `a **The recommended and default mode is `fork`. Only use `async` if you know what you're doing and have strong reasons to** +This supervisor mode is separate from a worker's concurrency model. Supervisor mode decides whether supervised processes live in forks or threads. Worker configuration decides whether claimed jobs run in a thread pool (`threads: N`) or as fibers on a single fiber reactor thread (`fibers: N`). + +Because these are separate concerns, you can combine the default `fork` supervisor mode with fiber workers. In that setup, each worker process gets its own fiber reactor and bounded fiber count. + ## Configuration By default, Solid Queue will try to find your configuration under `config/queue.yml`, but you can set a different path using the environment variable `SOLID_QUEUE_CONFIG` or by using the `-c/--config_file` option with `bin/jobs`, like this: @@ -223,9 +227,9 @@ production: batch_size: 500 concurrency_maintenance_interval: 300 workers: - - queues: "*" - threads: 3 - polling_interval: 2 + - queues: "llm*" + fibers: 100 + polling_interval: 0.05 - queues: [ real_time, background ] threads: 5 polling_interval: 0.1 @@ -272,9 +276,11 @@ Here's an overview of the different options: Check the sections below on [how queue order behaves combined with priorities](#queue-order-and-priorities), and [how the way you specify the queues per worker might affect performance](#queues-specification-and-performance). -- `threads`: this is the max size of the thread pool that each worker will have to run jobs. Each worker will fetch this number of jobs from their queue(s), at most and will post them to the thread pool to be run. By default, this is `3`. Only workers have this setting. -It is recommended to set this value less than or equal to the queue database's connection pool size minus 2, as each worker thread uses one connection, and two additional connections are reserved for polling and heartbeat. -- `processes`: this is the number of worker processes that will be forked by the supervisor with the settings given. By default, this is `1`, just a single process. This setting is useful if you want to dedicate more than one CPU core to a queue or queues with the same configuration. Only workers have this setting. **Note**: this option will be ignored if [running in `async` mode](#fork-vs-async-mode). +- `threads`: configures a worker to execute jobs in a thread pool of this size. By default, workers use `threads: 3`. Only workers have this setting, and it can't be combined with `fibers`. +It is recommended to set this value less than or equal to the queue database's connection pool size minus 2, as each worker uses connections for polling and heartbeat and thread mode may use additional connections for job execution. +- `fibers`: configures a worker to execute jobs as fibers on a single fiber reactor thread, with this value as the maximum number of in-flight jobs. It can't be combined with `threads`. + Fiber workers require fiber-scoped isolated execution state. In Rails apps, set `config.active_support.isolation_level = :fiber` before using `fibers`. Solid Queue refuses to boot fiber workers when isolation remains thread-scoped. On Rails 7.2 and later, a practical starting point is usually `3-5` queue database connections per worker process rather than matching the `fibers` value, because ordinary Active Record query paths can release connections between non-blocking waits. On Rails 7.1, size the queue database pool more conservatively, as in-flight fiber jobs may still retain connections roughly in proportion to `fibers`. +- `processes`: this is the number of worker processes that will be forked by the supervisor with the settings given. By default, this is `1`, just a single process. This setting is useful if you want to dedicate more than one CPU core to a queue or queues with the same configuration. Only workers have this setting. This works with both `threads` and `fibers` workers as long as the supervisor is running in the default `fork` mode. **Note**: this option is ignored only when the supervisor itself is [running in `async` mode](#fork-vs-async-mode). - `concurrency_maintenance`: whether the dispatcher will perform the concurrency maintenance work. This is `true` by default, and it's useful if you don't use any [concurrency controls](#concurrency-controls) and want to disable it or if you run multiple dispatchers and want some of them to just dispatch jobs without doing anything else. @@ -365,7 +371,15 @@ queues: back* ### Threads, processes, and signals -Workers in Solid Queue use a thread pool to run work in multiple threads, configurable via the `threads` parameter above. Besides this, parallelism can be achieved via multiple processes on one machine (configurable via different workers or the `processes` parameter above) or by horizontal scaling. +By default, workers in Solid Queue use a thread pool to run work in multiple threads, configurable via the `threads` parameter above. Workers can also be configured with `fibers`, in which case claimed jobs are executed as fibers on a single reactor thread and bounded by the worker's fiber count. Besides this, parallelism can be achieved via multiple processes on one machine (configurable via different workers or the `processes` parameter above) or by horizontal scaling. + +Fiber worker execution is best suited for cooperative, mostly I/O-bound jobs. Blocking or CPU-heavy work still blocks the single reactor thread, so it should not be expected to outperform thread mode for every workload. + +Because fiber workers run multiple fibers on a single thread, Rails must also isolate execution state per fiber rather than per thread. If your app keeps the default thread-scoped isolation level, Solid Queue will raise a boot-time error instead of running fiber workers with shared Active Record state. + +On Rails 7.2 and later, fiber workers can often use a much smaller queue database pool than an equivalent thread pool. A practical starting point is `3-5` queue database connections per worker process: one for job execution, one for polling, one for heartbeats, plus some headroom. In the default `fork` supervisor mode, that guidance applies per worker process. In supervisor `async` mode, all workers share one process, so add together the requirements for the workers running there. + +That lower-pool guidance depends on job code not holding connections open across non-blocking waits. APIs such as `ActiveRecord::Base.connection`, `lease_connection`, `connection_pool.checkout`, or long-lived `with_connection` / transaction blocks can pin connections and push fiber workers back toward thread-like pool usage. On Rails 7.1, plan conservatively and assume the configured fiber count can still grow queue database connection usage. The supervisor is in charge of managing these processes, and it responds to the following signals when running in its own process via `bin/jobs` or with [the Puma plugin](#puma-plugin) with the default `fork` mode: - `TERM`, `INT`: starts graceful termination. The supervisor will send a `TERM` signal to its supervised processes, and it'll wait up to `SolidQueue.shutdown_timeout` time until they're done. If any supervised processes are still around by then, it'll send a `QUIT` signal to them to indicate they must exit. @@ -377,6 +391,10 @@ On Windows, the `QUIT` signal can't be trapped, so the supervisor only responds If processes have no chance of cleaning up before exiting (e.g. if someone pulls a cable somewhere), in-flight jobs might remain claimed by the processes executing them. Processes send heartbeats, and the supervisor checks and prunes processes with expired heartbeats. Jobs that were claimed by processes with an expired heartbeat will be marked as failed with a `SolidQueue::Processes::ProcessPrunedError`. You can configure both the frequency of heartbeats and the threshold to consider a process dead. See the section below for this. +Worker heartbeats are driven by a separate timer task, not by the worker execution backend itself. This means fiber workers do not rely on the reactor loop to prove liveness. However, liveness is still tracked at the worker-process level, not at the individual thread or fiber level. + +This means finished and failed jobs still follow the normal Solid Queue lifecycle, but a single stuck job can remain claimed if the worker process itself is still alive. If you need stronger stuck-job detection, that requires an explicit timeout or watchdog mechanism on top of process heartbeats. + In a similar way, if a worker is terminated in any other way not initiated by the above signals (e.g. a worker is sent a `KILL` signal), jobs in progress will be marked as failed so that they can be inspected, with a `SolidQueue::Processes::ProcessExitError`. Sometimes a job in particular is responsible for this, for example, if it has a memory leak and you have a mechanism to kill processes over a certain memory threshold, so this will help identifying this kind of situation. @@ -392,7 +410,7 @@ _Note_: The settings in this section should be set in your `config/application.r There are several settings that control how Solid Queue works that you can set as well: - `logger`: the logger you want Solid Queue to use. Defaults to the app logger. -- `app_executor`: the [Rails executor](https://guides.rubyonrails.org/threading_and_code_execution.html#executor) used to wrap asynchronous operations, defaults to the app executor +- `app_executor`: the [Rails executor](https://guides.rubyonrails.org/threading_and_code_execution.html#executor) used to wrap background operations, defaults to the app executor - `on_thread_error`: custom lambda/Proc to call when there's an error within a Solid Queue thread that takes the exception raised as argument. Defaults to ```ruby diff --git a/lib/solid_queue/configuration.rb b/lib/solid_queue/configuration.rb index d04a0aaab..200245df9 100644 --- a/lib/solid_queue/configuration.rb +++ b/lib/solid_queue/configuration.rb @@ -6,7 +6,9 @@ class Configuration include ActiveModel::Validations::Callbacks validate :ensure_configured_processes, :ensure_valid_recurring_tasks - validate :warn_about_incorrectly_sized_thread_pool, :warn_about_missing_config_files + validate :ensure_valid_worker_execution_options + validate :ensure_fiber_workers_have_required_dependency, :ensure_fiber_workers_use_supported_isolation_level + validate :warn_about_incorrectly_sized_database_pool, :warn_about_missing_config_files before_validation { warnings.clear } @@ -37,6 +39,7 @@ def instantiate DEFAULT_CONFIG_FILE_PATH = "config/queue.yml" DEFAULT_RECURRING_SCHEDULE_FILE_PATH = "config/recurring.yml" + FIBER_QUERY_SCOPED_CONNECTIONS_VERSION = Gem::Version.new("7.2.0") def initialize(**options) @options = options.with_defaults(default_options) @@ -96,12 +99,12 @@ def ensure_valid_recurring_tasks end end - def warn_about_incorrectly_sized_thread_pool + def warn_about_incorrectly_sized_database_pool db_pool_size = SolidQueue::Record.connection_pool&.size - if db_pool_size && db_pool_size < estimated_number_of_threads - warnings.add(:base, "Warning: Solid Queue is configured to use #{estimated_number_of_threads} threads but the " \ - "database connection pool is #{db_pool_size}. Increase it in `config/database.yml`") + if db_pool_size && db_pool_size < estimated_database_pool_size + warnings.add(:base, "Warning: Solid Queue needs at least #{estimated_database_pool_size} database connections " \ + "for the configured workers but the database connection pool is #{db_pool_size}. Increase it in `config/database.yml`") end rescue ActiveRecord::ActiveRecordError # No usable database connection. Skip the pool-size warning in that case. @@ -118,6 +121,30 @@ def warn_about_missing_config_files end end + def ensure_valid_worker_execution_options + workers_options.each do |options| + if options.key?(:threads) && options.key?(:fibers) + errors.add(:base, "Workers can specify either `threads` or `fibers`, but not both.") + end + end + end + + def ensure_fiber_workers_have_required_dependency + return unless workers_options.any? { |options| fiber_worker?(options) } + + SolidQueue::ExecutionPools::FiberPool.ensure_dependency! + rescue LoadError => error + errors.add(:base, error.message) + end + + def ensure_fiber_workers_use_supported_isolation_level + return unless workers_options.any? { |options| fiber_worker?(options) } + + SolidQueue::ExecutionPools::FiberPool.ensure_supported_isolation_level! + rescue ArgumentError => error + errors.add(:base, error.message) + end + def default_options { mode: ENV["SOLID_QUEUE_SUPERVISOR_MODE"] || :fork, @@ -154,7 +181,8 @@ def workers 1 end - processes.times.map { Process.new(:worker, worker_options.with_defaults(WORKER_DEFAULTS)) } + defaults = worker_defaults_for(worker_options) + processes.times.map { Process.new(:worker, worker_options.with_defaults(defaults)) } end end @@ -248,10 +276,42 @@ def load_config_from_file(file) end end - def estimated_number_of_threads - # At most "threads" in each worker + 1 thread for the worker + 1 thread for the heartbeat task - thread_count = workers_options.map { |options| options.fetch(:threads, WORKER_DEFAULTS[:threads]) }.max - (thread_count || 1) + 2 + def estimated_database_pool_size + worker_pool_size = workers_options.map { |options| estimated_database_pool_size_for_worker(options) }.max + worker_pool_size || 1 + end + + def estimated_database_pool_size_for_worker(options) + # Connections used to execute jobs + 1 for the worker's polling thread + 1 for the heartbeat task + estimated_execution_connections_for_worker(options) + 2 + end + + def worker_capacity(options) + options[:fibers] || options[:threads] || WORKER_DEFAULTS[:threads] + end + + def estimated_execution_connections_for_worker(options) + fiber_worker?(options) ? fiber_execution_connections_for_worker(options) : worker_capacity(options) + end + + def fiber_execution_connections_for_worker(options) + fiber_jobs_release_connections_between_queries? ? 1 : worker_capacity(options) + end + + def fiber_jobs_release_connections_between_queries? + ActiveRecord.gem_version >= FIBER_QUERY_SCOPED_CONNECTIONS_VERSION + end + + def fiber_worker?(options) + options.key?(:fibers) + end + + def worker_defaults_for(options) + if fiber_worker?(options) + WORKER_DEFAULTS.except(:threads) + else + WORKER_DEFAULTS + end end end end diff --git a/lib/solid_queue/execution_pools.rb b/lib/solid_queue/execution_pools.rb new file mode 100644 index 000000000..766cc1894 --- /dev/null +++ b/lib/solid_queue/execution_pools.rb @@ -0,0 +1,18 @@ +# frozen_string_literal: true + +module SolidQueue + module ExecutionPools + class << self + def build(type:, size:, on_state_change: nil) + case type + when :thread + ThreadPool.new(size, on_state_change: on_state_change) + when :fiber + FiberPool.new(size, on_state_change: on_state_change) + else + raise ArgumentError, "Unknown execution pool type #{type.inspect}. Expected one of: :thread, :fiber" + end + end + end + end +end diff --git a/lib/solid_queue/execution_pools/fiber_pool.rb b/lib/solid_queue/execution_pools/fiber_pool.rb new file mode 100644 index 000000000..3234e3430 --- /dev/null +++ b/lib/solid_queue/execution_pools/fiber_pool.rb @@ -0,0 +1,239 @@ +# frozen_string_literal: true + +module SolidQueue + module ExecutionPools + class FiberPool + include AppExecutor + + IDLE_WAIT_INTERVAL = 0.01 + + class MissingDependencyError < LoadError + def initialize(error) + super( + "Fiber worker execution requires the `async` gem. " \ + "Add `gem \"async\"` to your Gemfile to configure workers with `fibers`. " \ + "Original error: #{error.message}" + ) + end + end + + class UnsupportedIsolationLevelError < ArgumentError + def initialize(level) + super( + "Fiber worker execution requires fiber-scoped isolated execution state. " \ + "Set `ActiveSupport::IsolatedExecutionState.isolation_level = :fiber` " \ + "(or `config.active_support.isolation_level = :fiber` in Rails). " \ + "Current isolation level: #{level.inspect}" + ) + end + end + + class << self + def ensure_dependency! + ensure_io_timeout_compatibility! + + require "async" + require "async/semaphore" + rescue LoadError => error + raise MissingDependencyError.new(error) + end + + def ensure_supported_isolation_level! + return if supported_isolation_level? + + raise UnsupportedIsolationLevelError.new(ActiveSupport::IsolatedExecutionState.isolation_level) + end + + def supported_isolation_level? + ActiveSupport::IsolatedExecutionState.isolation_level == :fiber + end + + def ensure_io_timeout_compatibility!(io_class = IO) + unless io_class.method_defined?(:timeout) && io_class.method_defined?(:timeout=) + # Async 2.24, which Ruby 3.1 resolves to, expects Ruby's newer IO + # timeout API to exist on any socket it waits on. Older Rubies don't + # provide it, so give the gem the minimal accessor interface it needs. + io_class.class_eval do + def timeout + @timeout + end + + def timeout=(value) + @timeout = value + end + end + end + + return if io_class.const_defined?(:TimeoutError, false) + + io_class.const_set(:TimeoutError, Class.new(StandardError)) + end + end + + attr_reader :size + + def initialize(size, on_state_change: nil) + @size = size + @on_state_change = on_state_change + @available_capacity = size + @mutex = Mutex.new + @state_mutex = Mutex.new + @shutdown = false + @fatal_error = nil + @boot_queue = Thread::Queue.new + @pending_executions = Thread::Queue.new + + self.class.ensure_dependency! + self.class.ensure_supported_isolation_level! + + @reactor_thread = start_reactor + + boot_result = @boot_queue.pop + raise boot_result if boot_result.is_a?(Exception) + end + + def post(execution) + reserved = false + raise_if_fatal_error! + raise RuntimeError, "Execution pool is shutting down" if shutdown? + + reserve_capacity! + reserved = true + pending_executions << execution + rescue Exception + restore_capacity if reserved + raise + end + + def available_capacity + raise_if_fatal_error! + mutex.synchronize { @available_capacity } + end + + def idle? + available_capacity.positive? + end + + def shutdown + state_mutex.synchronize do + next false if @shutdown + + @shutdown = true + end + end + + def shutdown? + state_mutex.synchronize { @shutdown } + end + + def wait_for_termination(timeout) + reactor_thread.join(timeout) + end + + def metadata + { + fiber_pool_size: size, + inflight: size - available_capacity + } + end + + private + attr_reader :boot_queue, :mutex, :on_state_change, :pending_executions, :reactor_thread, :state_mutex + + def name + @name ||= "solid_queue-fiber-pool-#{object_id}" + end + + def start_reactor + create_thread do + Async do |task| + semaphore = Async::Semaphore.new(size, parent: task) + boot_queue << :ready + + wait_for_executions(semaphore) + wait_for_inflight_executions + end + rescue Exception => error + register_fatal_error(error) + raise + end + end + + def wait_for_executions(semaphore) + loop do + schedule_pending_executions(semaphore) + + break if shutdown? && pending_executions.empty? + + # Older versions of the async gem don't support waking the reactor from another + # thread reliably, so we cooperatively poll for newly posted work. + sleep(IDLE_WAIT_INTERVAL) if pending_executions.empty? + end + end + + def schedule_pending_executions(semaphore) + while execution = next_pending_execution + semaphore.async(execution) do |_execution_task, scheduled_execution| + perform_execution(scheduled_execution) + end + end + end + + def next_pending_execution + pending_executions.pop(true) + rescue ThreadError + nil + end + + def perform_execution(execution) + wrap_in_app_executor { execution.perform } + rescue Async::Stop => error + handle_thread_error(error) + register_fatal_error(error) + rescue Exception => error + handle_thread_error(error) + ensure + restore_capacity + end + + def reserve_capacity! + mutex.synchronize do + raise RuntimeError, "Execution pool is at capacity" if @available_capacity <= 0 + + @available_capacity -= 1 + end + end + + def restore_capacity + should_notify = mutex.synchronize do + @available_capacity += 1 + @available_capacity.positive? + end + + on_state_change&.call if should_notify + end + + def register_fatal_error(error) + state_mutex.synchronize do + @fatal_error ||= error + end + + boot_queue << error if boot_queue.empty? + on_state_change&.call + end + + def raise_if_fatal_error! + error = state_mutex.synchronize { @fatal_error } + raise error if error + end + + def wait_for_inflight_executions + sleep(IDLE_WAIT_INTERVAL) while executions_in_flight? + end + + def executions_in_flight? + mutex.synchronize { @available_capacity < size } + end + end + end +end diff --git a/lib/solid_queue/execution_pools/thread_pool.rb b/lib/solid_queue/execution_pools/thread_pool.rb new file mode 100644 index 000000000..5b7d15653 --- /dev/null +++ b/lib/solid_queue/execution_pools/thread_pool.rb @@ -0,0 +1,82 @@ +# frozen_string_literal: true + +module SolidQueue + module ExecutionPools + class ThreadPool + include AppExecutor + + attr_reader :size + + delegate :shutdown, :shutdown?, :wait_for_termination, to: :executor + + def initialize(size, on_state_change: nil) + @size = size + @on_state_change = on_state_change + @available_capacity = size + @mutex = Mutex.new + end + + def post(execution) + reserved = false + reserve_capacity! + reserved = true + + Concurrent::Promises.future_on(executor, execution) do |thread_execution| + wrap_in_app_executor { thread_execution.perform } + rescue Exception => error + handle_thread_error(error) + ensure + restore_capacity + end + rescue Exception + restore_capacity if reserved + raise + end + + def available_capacity + mutex.synchronize { @available_capacity } + end + + def idle? + available_capacity.positive? + end + + def metadata + { + inflight: size - available_capacity, + thread_pool_size: size + } + end + + private + attr_reader :mutex, :on_state_change + + DEFAULT_OPTIONS = { + min_threads: 0, + idletime: 60, + fallback_policy: :abort + } + + def executor + @executor ||= Concurrent::ThreadPoolExecutor.new DEFAULT_OPTIONS.merge(max_threads: size, max_queue: size) + end + + def reserve_capacity! + mutex.synchronize do + raise RuntimeError, "Execution pool is at capacity" if @available_capacity <= 0 + + @available_capacity -= 1 + end + end + + def restore_capacity + should_notify = mutex.synchronize do + @available_capacity += 1 + @available_capacity.positive? + end + + on_state_change&.call if should_notify + end + end + end +end diff --git a/lib/solid_queue/pool.rb b/lib/solid_queue/pool.rb index 9c3d2a298..356bbaf35 100644 --- a/lib/solid_queue/pool.rb +++ b/lib/solid_queue/pool.rb @@ -1,54 +1,5 @@ # frozen_string_literal: true module SolidQueue - class Pool - include AppExecutor - - attr_reader :size - - delegate :shutdown, :shutdown?, :wait_for_termination, to: :executor - - def initialize(size, on_idle: nil) - @size = size - @on_idle = on_idle - @available_threads = Concurrent::AtomicFixnum.new(size) - @mutex = Mutex.new - end - - def post(execution) - available_threads.decrement - - Concurrent::Promises.future_on(executor, execution) do |thread_execution| - wrap_in_app_executor do - thread_execution.perform - ensure - available_threads.increment - mutex.synchronize { on_idle.try(:call) if idle? } - end - end.on_rejection! do |e| - handle_thread_error(e) - end - end - - def idle_threads - available_threads.value - end - - def idle? - idle_threads > 0 - end - - private - attr_reader :available_threads, :on_idle, :mutex - - DEFAULT_OPTIONS = { - min_threads: 0, - idletime: 60, - fallback_policy: :abort - } - - def executor - @executor ||= Concurrent::ThreadPoolExecutor.new DEFAULT_OPTIONS.merge(max_threads: size, max_queue: size) - end - end + Pool = ExecutionPools::ThreadPool end diff --git a/lib/solid_queue/worker.rb b/lib/solid_queue/worker.rb index e036a5fd9..31945b95b 100644 --- a/lib/solid_queue/worker.rb +++ b/lib/solid_queue/worker.rb @@ -11,18 +11,29 @@ class Worker < Processes::Poller attr_reader :queues, :pool def initialize(**options) - options = options.dup.with_defaults(SolidQueue::Configuration::WORKER_DEFAULTS) + options = options.dup + validate_execution_options!(options) + + execution_pool_type = options.key?(:fibers) ? :fiber : :thread + execution_pool_size = options[:fibers] || options[:threads] || SolidQueue::Configuration::WORKER_DEFAULTS[:threads] + options = options.with_defaults(worker_defaults_for(options)) # Ensure that the queues array is deep frozen to prevent accidental modification @queues = Array(options[:queues]).map(&:freeze).freeze + @metadata_state_mutex = Mutex.new + @metadata_dirty = false - @pool = Pool.new(options[:threads], on_idle: -> { wake_up }) + @pool_options = { + type: execution_pool_type, + size: execution_pool_size, + on_state_change: -> { mark_metadata_dirty; wake_up } + } super(**options) end def metadata - super.merge(queues: queues.join(","), thread_pool_size: pool.size) + super.merge(queues: queues.join(",")).merge(pool.metadata) end private @@ -32,16 +43,23 @@ def poll pool.post(execution) end + reload_metadata_if_needed(executions.any?) + pool.idle? ? polling_interval : 10.minutes end end def claim_executions with_polling_volume do - SolidQueue::ReadyExecution.claim(queues, pool.idle_threads, process_id) + SolidQueue::ReadyExecution.claim(queues, pool.available_capacity, process_id) end end + def boot + build_pool + super + end + def shutdown pool.shutdown pool.wait_for_termination(SolidQueue.shutdown_timeout) @@ -53,8 +71,49 @@ def all_work_completed? SolidQueue::ReadyExecution.aggregated_count_across(queues).zero? end + def heartbeat + super + reload_metadata + end + def set_procline procline "waiting for jobs in #{queues.join(",")}" end + + def build_pool + @pool ||= ExecutionPools.build(**@pool_options) + end + + def validate_execution_options!(options) + if options.key?(:threads) && options.key?(:fibers) + raise ArgumentError, "Workers can specify either `threads` or `fibers`, but not both." + end + end + + def worker_defaults_for(options) + if options.key?(:fibers) + SolidQueue::Configuration::WORKER_DEFAULTS.except(:threads) + else + SolidQueue::Configuration::WORKER_DEFAULTS + end + end + + def mark_metadata_dirty + metadata_state_mutex.synchronize { @metadata_dirty = true } + end + + def metadata_state_mutex + @metadata_state_mutex + end + + def reload_metadata_if_needed(executions_claimed) + needs_reload = metadata_state_mutex.synchronize do + claimed_or_dirty = executions_claimed || @metadata_dirty + @metadata_dirty = false + claimed_or_dirty + end + + reload_metadata if needs_reload + end end end diff --git a/solid_queue.gemspec b/solid_queue.gemspec index 425a5adc6..19172043d 100644 --- a/solid_queue.gemspec +++ b/solid_queue.gemspec @@ -34,6 +34,7 @@ Gem::Specification.new do |spec| spec.add_development_dependency "appraisal" spec.add_development_dependency "debug", "~> 1.9" + spec.add_development_dependency "async", ">= 2.24" spec.add_development_dependency "minitest", "~> 5.0" spec.add_development_dependency "mocha" spec.add_development_dependency "puma", "~> 7.0" diff --git a/test/integration/async_processes_lifecycle_test.rb b/test/integration/async_processes_lifecycle_test.rb index 6f269e220..dc56440b6 100644 --- a/test/integration/async_processes_lifecycle_test.rb +++ b/test/integration/async_processes_lifecycle_test.rb @@ -125,10 +125,16 @@ class AsyncProcessesLifecycleTest < ActiveSupport::TestCase no_pause = enqueue_store_result_job("no pause") pause = enqueue_store_result_job("pause", pause: SolidQueue.shutdown_timeout + 10.second) - # Wait for the "no pause" job to complete and the pause job to be claimed. - # This ensures the pause job is actively being processed. + # Wait for the "no pause" job to complete and the pause job to start. + # A claimed execution alone is not enough here because the worker may have + # claimed the job but not yet entered `perform`. wait_for_jobs_to_finish_for(3.seconds, except: pause) - wait_for(timeout: 2.seconds) { SolidQueue::ClaimedExecution.exists?(job_id: SolidQueue::Job.find_by(active_job_id: pause.job_id)&.id) } + wait_for(timeout: 5.seconds) do + skip_active_record_query_cache do + JobResult.where(queue_name: :background, status: "started", value: "pause").exists? + end + end + assert_started_job_result("pause") signal_process(@pid, :TERM, wait: 0.2.second) wait_for_jobs_to_finish_for(2.seconds, except: pause) @@ -142,8 +148,7 @@ class AsyncProcessesLifecycleTest < ActiveSupport::TestCase assert_completed_job_results("no pause") assert_job_status(no_pause, :finished) - # The pause job should have started but not completed - assert_started_job_result("pause") + # The pause job should not have completed assert_not_equal "completed", skip_active_record_query_cache { JobResult.find_by(value: "pause")&.status } # After shutdown, the pause job may be either: diff --git a/test/integration/instrumentation_test.rb b/test/integration/instrumentation_test.rb index 1822cf159..1817f326b 100644 --- a/test/integration/instrumentation_test.rb +++ b/test/integration/instrumentation_test.rb @@ -10,7 +10,7 @@ class InstrumentationTest < ActiveSupport::TestCase travel_to 2.days.from_now dispatcher = SolidQueue::Dispatcher.new(polling_interval: 0.1, batch_size: 10).tap(&:start) - wait_while_with_timeout!(0.5.seconds) { SolidQueue::ScheduledExecution.any? } + wait_while_with_timeout!(3.seconds) { SolidQueue::ScheduledExecution.any? } dispatcher.stop end diff --git a/test/test_helper.rb b/test/test_helper.rb index 99d00ba27..ca90ddbce 100644 --- a/test/test_helper.rb +++ b/test/test_helper.rb @@ -29,6 +29,19 @@ def write(...) Logger::LogDevice.prepend(BlockLogDeviceTimeoutExceptions) class ExpectedTestError < RuntimeError; end +module ExecutionIsolationTestHelper + def with_execution_isolation(level) + previous_level = ActiveSupport::IsolatedExecutionState.isolation_level + ActiveSupport::IsolatedExecutionState.isolation_level = level + yield + ensure + ActiveSupport::IsolatedExecutionState.isolation_level = previous_level + end +end + +class Minitest::Test + include ExecutionIsolationTestHelper +end class ActiveSupport::TestCase include ConfigurationTestHelper, ProcessesTestHelper, JobsTestHelper diff --git a/test/unit/async_supervisor_test.rb b/test/unit/async_supervisor_test.rb index 962c4de7e..0ca6a3c2f 100644 --- a/test/unit/async_supervisor_test.rb +++ b/test/unit/async_supervisor_test.rb @@ -97,7 +97,7 @@ class AsyncSupervisorTest < ActiveSupport::TestCase supervisor.stop end - assert_match /Solid Queue is configured to use \d+ threads but the database connection pool is \d+\. Increase it in `config\/database.yml`/, log.string + assert_match /Solid Queue needs at least \d+ database connections for the configured workers but the database connection pool is \d+\. Increase it in `config\/database.yml`/, log.string end test "does not warn on boot when the database connection pool is large enough" do diff --git a/test/unit/configuration_test.rb b/test/unit/configuration_test.rb index 7bb7f70a5..b25bc0667 100644 --- a/test/unit/configuration_test.rb +++ b/test/unit/configuration_test.rb @@ -62,6 +62,65 @@ class ConfigurationTest < ActiveSupport::TestCase assert_processes configuration, :worker, 2 end + test "fibers configure workers to execute jobs in fibers" do + with_execution_isolation(:fiber) do + configuration = SolidQueue::Configuration.new( + workers: [ + { queues: "llm*", fibers: 10 }, + { queues: "*", threads: 3 } + ], + dispatchers: [], + skip_recurring: true + ) + + assert configuration.valid? + assert_processes configuration, :worker, 2, fibers: [ 10, nil ], threads: [ nil, 3 ] + end + end + + test "workers reject threads and fibers together" do + with_execution_isolation(:fiber) do + configuration = SolidQueue::Configuration.new( + workers: [ { queues: "llm*", threads: 1, fibers: 1 } ], + dispatchers: [], + skip_recurring: true + ) + + assert_not configuration.valid? + assert_match /either `threads` or `fibers`/, configuration.errors.full_messages.first + end + end + + test "fiber worker size inflates required database pool size on Rails 7.1" do + skip if fiber_workers_release_connections_between_queries? + + with_execution_isolation(:fiber) do + configuration = SolidQueue::Configuration.new( + workers: [ { queues: "llm*", fibers: 1000 } ], + dispatchers: [], + skip_recurring: true + ) + + assert configuration.valid? + assert_match /needs at least 1002 database connections/, configuration.warnings.full_messages.first + end + end + + test "fiber worker size does not inflate required database pool size on Rails 7.2+" do + skip unless fiber_workers_release_connections_between_queries? + + with_execution_isolation(:fiber) do + configuration = SolidQueue::Configuration.new( + workers: [ { queues: "llm*", fibers: 1000 } ], + dispatchers: [], + skip_recurring: true + ) + + assert configuration.valid? + assert_empty configuration.warnings + end + end + test "mulitple workers with the same configuration" do background_worker = { queues: "background", polling_interval: 10, processes: 3 } configuration = SolidQueue::Configuration.new(workers: [ background_worker ]) @@ -166,17 +225,33 @@ class ConfigurationTest < ActiveSupport::TestCase assert_not configuration.valid? assert_equal [ "No processes configured" ], configuration.errors.full_messages + # Fiber workers require fiber-scoped isolated execution state + configuration = SolidQueue::Configuration.new(skip_recurring: true, dispatchers: [], workers: [ { fibers: 3 } ]) + assert_not configuration.valid? + assert_match /requires fiber-scoped isolated execution state/, configuration.errors.full_messages.first + + # Fiber workers require the async gem + with_execution_isolation(:fiber) do + load_error = LoadError.new("cannot load such file -- async") + missing_dependency_error = SolidQueue::ExecutionPools::FiberPool::MissingDependencyError.new(load_error) + SolidQueue::ExecutionPools::FiberPool.expects(:ensure_dependency!).raises(missing_dependency_error) + + configuration = SolidQueue::Configuration.new(skip_recurring: true, dispatchers: [], workers: [ { fibers: 3 } ]) + assert_not configuration.valid? + assert_match /gem "async"/, configuration.errors.full_messages.first + end + # Not enough DB connections: still valid so boot is not blocked configuration = SolidQueue::Configuration.new(workers: [ { queues: "background", threads: 50, polling_interval: 10 } ]) assert configuration.valid? end - test "reports an undersized thread pool as a warning rather than an error" do + test "reports an undersized database pool as a warning rather than an error" do configuration = SolidQueue::Configuration.new(workers: [ { queues: "background", threads: 50, polling_interval: 10 } ], skip_recurring: true) assert configuration.valid? assert_equal 1, configuration.warnings.size - assert_match /Solid Queue is configured to use \d+ threads but the database connection pool is \d+\. Increase it in `config\/database.yml`/, configuration.warnings.full_messages.first + assert_match /Solid Queue needs at least \d+ database connections for the configured workers but the database connection pool is \d+\. Increase it in `config\/database.yml`/, configuration.warnings.full_messages.first end test "has no warnings when the database connection pool is large enough" do @@ -209,7 +284,7 @@ class ConfigurationTest < ActiveSupport::TestCase end assert_match "Solid Queue configuration is valid.", out - assert_match /Solid Queue is configured to use \d+ threads but the database connection pool is \d+/, err + assert_match /Solid Queue needs at least \d+ database connections for the configured workers but the database connection pool is \d+/, err end test "check prints errors to stderr and returns false for an invalid configuration" do @@ -222,6 +297,10 @@ class ConfigurationTest < ActiveSupport::TestCase end private + def fiber_workers_release_connections_between_queries? + ActiveRecord.gem_version >= SolidQueue::Configuration::FIBER_QUERY_SCOPED_CONNECTIONS_VERSION + end + def assert_processes(configuration, kind, count, **attributes) processes = configuration.configured_processes.select { |p| p.kind == kind } assert_equal count, processes.size diff --git a/test/unit/execution_pools/fiber_pool_test.rb b/test/unit/execution_pools/fiber_pool_test.rb new file mode 100644 index 000000000..668c00979 --- /dev/null +++ b/test/unit/execution_pools/fiber_pool_test.rb @@ -0,0 +1,135 @@ +require "test_helper" + +class FiberPoolTest < Minitest::Test + Execution = Struct.new(:started, :results, :pause) do + def perform + started << true if started + sleep(pause) if pause + results << [ Thread.current.object_id, Fiber.current.object_id ] if results + end + end + + CancelledExecution = Struct.new(:started) do + def perform + started << true if started + raise Async::Stop.new + end + end + + def test_raises_a_clear_error_when_the_async_gem_is_unavailable + load_error = LoadError.new("cannot load such file -- async") + + SolidQueue::ExecutionPools::FiberPool.expects(:require).with("async").raises(load_error) + + error = assert_raises SolidQueue::ExecutionPools::FiberPool::MissingDependencyError do + SolidQueue::ExecutionPools::FiberPool.new(3) + end + + assert_match /gem "async"/, error.message + end + + def test_builds_a_fiber_pool + pool = mock + + SolidQueue::ExecutionPools::FiberPool.expects(:new).with(5, on_state_change: nil).returns(pool) + + assert_equal pool, SolidQueue::ExecutionPools.build(type: :fiber, size: 5) + end + + def test_raises_a_clear_error_when_isolation_level_is_not_fiber + error = assert_raises SolidQueue::ExecutionPools::FiberPool::UnsupportedIsolationLevelError do + SolidQueue::ExecutionPools::FiberPool.new(3) + end + + assert_match /isolation_level = :fiber/, error.message + end + + def test_adds_io_timeout_compatibility_for_older_rubies + io_class = Class.new + + SolidQueue::ExecutionPools::FiberPool.ensure_io_timeout_compatibility!(io_class) + + io = io_class.new + assert_nil io.timeout + + io.timeout = 1.second + + assert_equal 1.second, io.timeout + assert io_class.const_defined?(:TimeoutError, false) + end + + def test_executes_jobs_as_fibers_on_a_single_reactor_thread + with_execution_isolation(:fiber) do + pool = SolidQueue::ExecutionPools::FiberPool.new(2) + results = Thread::Queue.new + + pool.post Execution.new(nil, results, 0.05) + pool.post Execution.new(nil, results, 0.05) + + entries = 2.times.map { Timeout.timeout(1.second) { results.pop } } + + assert_equal 1, entries.map(&:first).uniq.count + assert_equal 2, entries.map(&:last).uniq.count + assert_equal 2, pool.available_capacity + assert_equal 0, pool.metadata[:inflight] + ensure + pool&.shutdown + pool&.wait_for_termination(1.second) + end + end + + def test_waits_for_in_flight_executions_during_shutdown + with_execution_isolation(:fiber) do + pool = SolidQueue::ExecutionPools::FiberPool.new(1) + started = Thread::Queue.new + + pool.post Execution.new(started, nil, 0.1) + Timeout.timeout(1.second) { started.pop } + + pool.shutdown + + assert_nil pool.wait_for_termination(0.01) + assert pool.wait_for_termination(1.second) + ensure + pool&.shutdown + pool&.wait_for_termination(1.second) + end + end + + def test_shutdown_wakes_the_reactor_when_idle + with_execution_isolation(:fiber) do + pool = SolidQueue::ExecutionPools::FiberPool.new(1) + + pool.shutdown + + assert pool.wait_for_termination(1.second) + ensure + pool&.shutdown + pool&.wait_for_termination(1.second) + end + end + + def test_marks_the_pool_as_fatal_when_an_execution_is_cancelled + with_execution_isolation(:fiber) do + notifications = Thread::Queue.new + started = Thread::Queue.new + reported_errors = [] + original_on_thread_error = SolidQueue.on_thread_error + SolidQueue.on_thread_error = ->(error) { reported_errors << error.class.name } + + pool = SolidQueue::ExecutionPools::FiberPool.new(1, on_state_change: -> { notifications << :changed }) + + pool.post CancelledExecution.new(started) + Timeout.timeout(1.second) { started.pop } + Timeout.timeout(1.second) { notifications.pop } + + error = assert_raises(Async::Stop) { pool.available_capacity } + assert_equal [ error.class.name ], reported_errors + assert_raises(Async::Stop) { pool.metadata } + ensure + SolidQueue.on_thread_error = original_on_thread_error + pool&.shutdown + pool&.wait_for_termination(1.second) + end + end +end diff --git a/test/unit/worker_test.rb b/test/unit/worker_test.rb index 3d692404b..a7baecb25 100644 --- a/test/unit/worker_test.rb +++ b/test/unit/worker_test.rb @@ -3,6 +3,26 @@ class WorkerTest < ActiveSupport::TestCase include ActiveSupport::Testing::MethodCallAssertions + self.use_transactional_tests = false + + EXECUTION_MODES = [ + { + name: "thread", + options: { threads: 3 }, + expected_metadata: { + inflight: 0, + thread_pool_size: 3 + } + }, + { + name: "fiber", + options: { fibers: 3 }, + expected_metadata: { + inflight: 0, + fiber_pool_size: 3 + } + } + ].freeze setup do @worker = SolidQueue::Worker.new(queues: "background", threads: 3, polling_interval: 0.2) @@ -13,13 +33,117 @@ class WorkerTest < ActiveSupport::TestCase JobBuffer.clear end - test "worker is registered as process" do - @worker.start + EXECUTION_MODES.each do |mode| + test "worker is registered as process in #{mode[:name]} mode" do + with_worker_execution_support(mode[:options]) do + worker = SolidQueue::Worker.new(queues: "background", polling_interval: 0.2, **mode[:options]) + + worker.start + wait_for_registered_processes(1, timeout: 1.second) + + process = SolidQueue::Process.first + assert_equal "Worker", process.kind + assert_metadata process, { + queues: "background", + polling_interval: 0.2 + }.merge(mode[:expected_metadata]) + ensure + worker&.stop + wait_for_registered_processes(0, timeout: 1.second) + end + end + + test "claim and process more enqueued jobs than the pool size allows to process at once in #{mode[:name]} mode" do + 5.times do + StoreResultJob.perform_later(:paused, pause: 0.1.second) + end + + 3.times do + StoreResultJob.perform_later(:immediate) + end + + with_worker_execution_support(mode[:options]) do + worker = SolidQueue::Worker.new(queues: "background", polling_interval: 0.2, **mode[:options]) + + worker.start + + wait_for_jobs_to_finish_for(2.second) + worker.wake_up + + assert_equal 5, JobResult.where(queue_name: :background, status: "completed", value: :paused).count + assert_equal 3, JobResult.where(queue_name: :background, status: "completed", value: :immediate).count + ensure + worker&.stop + wait_for_registered_processes(0, timeout: 1.second) + end + end + + test "updates inflight metadata after jobs finish in #{mode[:name]} mode" do + StoreResultJob.perform_later(:slow, pause: 0.1.second) + + with_worker_execution_support(mode[:options]) do + worker = SolidQueue::Worker.new(queues: "background", polling_interval: 0.05, **mode[:options]) + + worker.start + wait_for_registered_processes(1, timeout: 1.second) + + process = SolidQueue::Process.first + + wait_for(timeout: 2.seconds) { process.reload.metadata["inflight"] == 1 } + assert_equal 1, process.reload.metadata["inflight"] + + wait_for(timeout: 2.seconds) do + process.reload.metadata["inflight"] == 0 && + JobResult.where(queue_name: :background, status: "completed", value: :slow).count == 1 + end + assert_equal 0, process.reload.metadata["inflight"] + assert_equal 1, JobResult.where(queue_name: :background, status: "completed", value: :slow).count + ensure + worker&.stop + wait_for_registered_processes(0, timeout: 1.second) + end + end + end + + test "builds the execution pool on boot instead of initialize" do + pool = SolidQueue::ExecutionPools::ThreadPool.new(3) + + SolidQueue::ExecutionPools.expects(:build).once.with do |**options| + options[:type] == :thread && options[:size] == 3 && options[:on_state_change].respond_to?(:call) + end.returns(pool) + + worker = SolidQueue::Worker.new(queues: "background", threads: 3, polling_interval: 0.2) + + assert_nil worker.pool + + worker.start wait_for_registered_processes(1, timeout: 1.second) - process = SolidQueue::Process.first - assert_equal "Worker", process.kind - assert_metadata process, { queues: "background", polling_interval: 0.2, thread_pool_size: 3 } + assert_equal pool, worker.pool + ensure + worker&.stop + wait_for_registered_processes(0, timeout: 1.second) + end + + test "rejects multiple worker pool size options" do + error = assert_raises ArgumentError do + SolidQueue::Worker.new(queues: "background", threads: 3, fibers: 3, polling_interval: 0.2) + end + + assert_match /either `threads` or `fibers`/, error.message + end + + test "defaults thread workers to the configured thread pool size" do + worker = SolidQueue::Worker.new(queues: "background", polling_interval: 0.2) + + worker.start + wait_for_registered_processes(1, timeout: 1.second) + + assert_equal 3, worker.pool.size + assert_metadata SolidQueue::Process.first, thread_pool_size: 3, inflight: 0 + ensure + worker&.stop + wait_for_registered_processes(0, timeout: 1.second) end test "errors on polling are passed to on_thread_error and re-raised" do @@ -85,24 +209,6 @@ class WorkerTest < ActiveSupport::TestCase Rails.error.unsubscribe(subscriber) if Rails.error.respond_to?(:unsubscribe) end - test "claim and process more enqueued jobs than the pool size allows to process at once" do - 5.times do |i| - StoreResultJob.perform_later(:paused, pause: 0.1.second) - end - - 3.times do |i| - StoreResultJob.perform_later(:immediate) - end - - @worker.start - - wait_for_jobs_to_finish_for(2.second) - @worker.wake_up - - assert_equal 5, JobResult.where(queue_name: :background, status: "completed", value: :paused).count - assert_equal 3, JobResult.where(queue_name: :background, status: "completed", value: :immediate).count - end - test "polling queries are logged" do log = StringIO.new with_active_record_logger(ActiveSupport::Logger.new(log)) do @@ -173,28 +279,50 @@ class WorkerTest < ActiveSupport::TestCase end test "sleeps `10.minutes` if at capacity" do - 3.times { |i| StoreResultJob.perform_later(i, pause: 1.second) } + 3.times { |i| StoreResultJob.perform_later(i, pause: 5.seconds) } - @worker.expects(:interruptible_sleep).with(10.minutes).at_least_once - @worker.expects(:interruptible_sleep).with(@worker.polling_interval).never - @worker.expects(:handle_thread_error).never + delays = stub_interruptible_sleep(@worker) @worker.start - sleep 1.second + + first_delay = Timeout.timeout(1.second) { delays.pop } + + assert_equal 10.minutes, first_delay end test "sleeps `polling_interval` if worker not at capacity" do - 2.times { |i| StoreResultJob.perform_later(i, pause: 1.second) } + 2.times { |i| StoreResultJob.perform_later(i, pause: 5.seconds) } - @worker.expects(:interruptible_sleep).with(@worker.polling_interval).at_least_once - @worker.expects(:interruptible_sleep).with(10.minutes).never - @worker.expects(:handle_thread_error).never + delays = stub_interruptible_sleep(@worker) @worker.start - sleep 1.second + + first_delay = Timeout.timeout(1.second) { delays.pop } + + assert_equal @worker.polling_interval, first_delay end private + def stub_interruptible_sleep(worker) + delays = Thread::Queue.new + + worker.stubs(:handle_thread_error) + worker.define_singleton_method(:interruptible_sleep) do |delay| + delays << delay + sleep 0.01 + end + + delays + end + + def with_worker_execution_support(options, &block) + if options.key?(:fibers) + with_execution_isolation(:fiber, &block) + else + yield + end + end + def with_polling(silence:) old_silence_polling, SolidQueue.silence_polling = SolidQueue.silence_polling, silence yield From bbe6b77cea3f40d080229e77fac8af6abbf3bb65 Mon Sep 17 00:00:00 2001 From: Rosa Gutierrez Date: Wed, 22 Jul 2026 20:35:17 +0200 Subject: [PATCH 2/4] Remove IO timeout compatibility shim for Ruby 3.1 Now that the minimum supported Ruby version is 3.2, IO#timeout, IO#timeout= and IO::TimeoutError, which async expects, are always available natively, and we no longer need to patch them in. Co-Authored-By: Claude Opus 4.8 --- lib/solid_queue/execution_pools/fiber_pool.rb | 23 ------------------- test/unit/execution_pools/fiber_pool_test.rb | 14 ----------- 2 files changed, 37 deletions(-) diff --git a/lib/solid_queue/execution_pools/fiber_pool.rb b/lib/solid_queue/execution_pools/fiber_pool.rb index 3234e3430..f99d6014c 100644 --- a/lib/solid_queue/execution_pools/fiber_pool.rb +++ b/lib/solid_queue/execution_pools/fiber_pool.rb @@ -30,8 +30,6 @@ def initialize(level) class << self def ensure_dependency! - ensure_io_timeout_compatibility! - require "async" require "async/semaphore" rescue LoadError => error @@ -47,27 +45,6 @@ def ensure_supported_isolation_level! def supported_isolation_level? ActiveSupport::IsolatedExecutionState.isolation_level == :fiber end - - def ensure_io_timeout_compatibility!(io_class = IO) - unless io_class.method_defined?(:timeout) && io_class.method_defined?(:timeout=) - # Async 2.24, which Ruby 3.1 resolves to, expects Ruby's newer IO - # timeout API to exist on any socket it waits on. Older Rubies don't - # provide it, so give the gem the minimal accessor interface it needs. - io_class.class_eval do - def timeout - @timeout - end - - def timeout=(value) - @timeout = value - end - end - end - - return if io_class.const_defined?(:TimeoutError, false) - - io_class.const_set(:TimeoutError, Class.new(StandardError)) - end end attr_reader :size diff --git a/test/unit/execution_pools/fiber_pool_test.rb b/test/unit/execution_pools/fiber_pool_test.rb index 668c00979..16e8f75bb 100644 --- a/test/unit/execution_pools/fiber_pool_test.rb +++ b/test/unit/execution_pools/fiber_pool_test.rb @@ -44,20 +44,6 @@ def test_raises_a_clear_error_when_isolation_level_is_not_fiber assert_match /isolation_level = :fiber/, error.message end - def test_adds_io_timeout_compatibility_for_older_rubies - io_class = Class.new - - SolidQueue::ExecutionPools::FiberPool.ensure_io_timeout_compatibility!(io_class) - - io = io_class.new - assert_nil io.timeout - - io.timeout = 1.second - - assert_equal 1.second, io.timeout - assert io_class.const_defined?(:TimeoutError, false) - end - def test_executes_jobs_as_fibers_on_a_single_reactor_thread with_execution_isolation(:fiber) do pool = SolidQueue::ExecutionPools::FiberPool.new(2) From f98b4646ce8f02f62f2809d8fa84fa9f6453b42e Mon Sep 17 00:00:00 2001 From: Rosa Gutierrez Date: Wed, 22 Jul 2026 20:35:28 +0200 Subject: [PATCH 3/4] Block on the pending executions queue instead of polling the reactor The fiber pool's reactor looped with a 10ms sleep to check for newly posted work, and again while waiting for in-flight executions to finish during shutdown, on the assumption that the async gem can't reliably wake the reactor from another thread. It can: Thread::Queue#pop is fiber-scheduler-aware, so a fiber blocking on it suspends properly while other fibers run, and a push from the poller thread wakes the reactor through the scheduler (via Scheduler#unblock and the selector's cross-thread wakeup, reliable since io-event 1.3, which all supported async versions require). This removes the idle CPU cost of the busy-poll and the up-to-10ms latency between claiming a job and executing it. Shutdown now closes the queue, which wakes the blocked pop, drains any already-queued executions and then returns nil, ending the loop. The separate wait for in-flight executions is unnecessary: they run in fibers that are children of the reactor's top-level task, which already waits for all its children before exiting. Co-Authored-By: Claude Opus 4.8 --- lib/solid_queue/execution_pools/fiber_pool.rb | 41 +++++-------------- 1 file changed, 11 insertions(+), 30 deletions(-) diff --git a/lib/solid_queue/execution_pools/fiber_pool.rb b/lib/solid_queue/execution_pools/fiber_pool.rb index f99d6014c..72023d643 100644 --- a/lib/solid_queue/execution_pools/fiber_pool.rb +++ b/lib/solid_queue/execution_pools/fiber_pool.rb @@ -5,8 +5,6 @@ module ExecutionPools class FiberPool include AppExecutor - IDLE_WAIT_INTERVAL = 0.01 - class MissingDependencyError < LoadError def initialize(error) super( @@ -96,6 +94,10 @@ def shutdown next false if @shutdown @shutdown = true + end.tap do |shut_down| + # Wake the reactor: already-queued executions are drained before the + # blocked pop in +wait_for_executions+ returns nil + pending_executions.close if shut_down end end @@ -127,8 +129,9 @@ def start_reactor semaphore = Async::Semaphore.new(size, parent: task) boot_queue << :ready + # The reactor exits when all in-flight execution fibers, children + # of this task, have finished wait_for_executions(semaphore) - wait_for_inflight_executions end rescue Exception => error register_fatal_error(error) @@ -137,31 +140,17 @@ def start_reactor end def wait_for_executions(semaphore) - loop do - schedule_pending_executions(semaphore) - - break if shutdown? && pending_executions.empty? - - # Older versions of the async gem don't support waking the reactor from another - # thread reliably, so we cooperatively poll for newly posted work. - sleep(IDLE_WAIT_INTERVAL) if pending_executions.empty? - end - end - - def schedule_pending_executions(semaphore) - while execution = next_pending_execution + # Thread::Queue#pop is fiber-scheduler-aware: it suspends this fiber, letting + # execution fibers run, and wakes the reactor when the poller thread pushes new + # work or closes the queue on shutdown, after which it drains any remaining + # executions and returns nil + while execution = pending_executions.pop semaphore.async(execution) do |_execution_task, scheduled_execution| perform_execution(scheduled_execution) end end end - def next_pending_execution - pending_executions.pop(true) - rescue ThreadError - nil - end - def perform_execution(execution) wrap_in_app_executor { execution.perform } rescue Async::Stop => error @@ -203,14 +192,6 @@ def raise_if_fatal_error! error = state_mutex.synchronize { @fatal_error } raise error if error end - - def wait_for_inflight_executions - sleep(IDLE_WAIT_INTERVAL) while executions_in_flight? - end - - def executions_in_flight? - mutex.synchronize { @available_capacity < size } - end end end end From 3559d3e01a4ce1ae2d9e428644da9613c6142d04 Mon Sep 17 00:00:00 2001 From: Rosa Gutierrez Date: Wed, 22 Jul 2026 20:59:59 +0200 Subject: [PATCH 4/4] Add integration test for fiber workers and clarify their docs Cover the full lifecycle of a supervised, forked fiber worker: jobs executing in fibers, graceful termination draining in-flight jobs, and abrupt termination releasing claimed jobs. The unit tests exercise the fiber pool in-process only, which wouldn't catch a pool built before forking, whose reactor thread wouldn't survive the fork. Also restore the generic first worker in the configuration example, adding the fiber worker as an extra entry instead of replacing it, and note that the fiber isolation level is an application-wide setting, not scoped to Solid Queue's processes. Co-Authored-By: Claude Opus 4.8 --- README.md | 11 +- .../fiber_processes_lifecycle_test.rb | 124 ++++++++++++++++++ 2 files changed, 132 insertions(+), 3 deletions(-) create mode 100644 test/integration/fiber_processes_lifecycle_test.rb diff --git a/README.md b/README.md index 005b5ea5c..997270672 100644 --- a/README.md +++ b/README.md @@ -227,13 +227,16 @@ production: batch_size: 500 concurrency_maintenance_interval: 300 workers: - - queues: "llm*" - fibers: 100 - polling_interval: 0.05 + - queues: "*" + threads: 3 + polling_interval: 2 - queues: [ real_time, background ] threads: 5 polling_interval: 0.1 processes: 3 + - queues: "api*" + fibers: 100 + polling_interval: 0.05 scheduler: dynamic_tasks_enabled: true polling_interval: 5 @@ -377,6 +380,8 @@ Fiber worker execution is best suited for cooperative, mostly I/O-bound jobs. Bl Because fiber workers run multiple fibers on a single thread, Rails must also isolate execution state per fiber rather than per thread. If your app keeps the default thread-scoped isolation level, Solid Queue will raise a boot-time error instead of running fiber workers with shared Active Record state. +Keep in mind that `config.active_support.isolation_level = :fiber` applies to your whole application, not just to Solid Queue: if you run Solid Queue inside Puma via [the plugin](#puma-plugin), or combine fiber workers with thread workers in the same process using the supervisor's `async` mode, everything in that process will use fiber-scoped execution state. This is fully supported by Rails, but it's a global setting worth being deliberate about. + On Rails 7.2 and later, fiber workers can often use a much smaller queue database pool than an equivalent thread pool. A practical starting point is `3-5` queue database connections per worker process: one for job execution, one for polling, one for heartbeats, plus some headroom. In the default `fork` supervisor mode, that guidance applies per worker process. In supervisor `async` mode, all workers share one process, so add together the requirements for the workers running there. That lower-pool guidance depends on job code not holding connections open across non-blocking waits. APIs such as `ActiveRecord::Base.connection`, `lease_connection`, `connection_pool.checkout`, or long-lived `with_connection` / transaction blocks can pin connections and push fiber workers back toward thread-like pool usage. On Rails 7.1, plan conservatively and assume the configured fiber count can still grow queue database connection usage. diff --git a/test/integration/fiber_processes_lifecycle_test.rb b/test/integration/fiber_processes_lifecycle_test.rb new file mode 100644 index 000000000..3a1eafe35 --- /dev/null +++ b/test/integration/fiber_processes_lifecycle_test.rb @@ -0,0 +1,124 @@ +# frozen_string_literal: true + +require "test_helper" + +class FiberProcessesLifecycleTest < ActiveSupport::TestCase + self.use_transactional_tests = false + + setup do + @pid = fork do + ActiveSupport::IsolatedExecutionState.isolation_level = :fiber + SolidQueue::Supervisor.start(skip_recurring: true, workers: [ { queues: :background, fibers: 3, polling_interval: 0.1 } ]) + end + + wait_for_registered_processes(2, timeout: 3.second) + assert_registered_workers_for(:background, supervisor_pid: @pid) + end + + teardown do + terminate_process(@pid) if process_exists?(@pid) + end + + test "process jobs in fibers in a forked worker" do + worker = find_processes_registered_as("Worker").first + assert_metadata worker, fiber_pool_size: 3 + + 6.times { |i| enqueue_store_result_job("job_#{i}") } + + wait_for_jobs_to_finish_for(3.seconds) + + assert_equal 6, skip_active_record_query_cache { JobResult.count } + 6.times { |i| assert_completed_job_results("job_#{i}", :background) } + + terminate_process(@pid) + assert_clean_termination + end + + test "term supervisor while there are jobs in-flight" do + no_pause = enqueue_store_result_job("no pause") + pause = enqueue_store_result_job("pause", pause: 1.second) + + signal_process(@pid, :TERM, wait: 0.3.second) + wait_for_jobs_to_finish_for(3.seconds) + + assert_completed_job_results("no pause") + assert_completed_job_results("pause") + + assert_job_status(no_pause, :finished) + assert_job_status(pause, :finished) + + wait_for_process_termination_with_timeout(@pid, timeout: 1.second) + assert_clean_termination + end + + test "quit supervisor while there are jobs in-flight" do + no_pause = enqueue_store_result_job("no pause") + # long enough pause to make sure it doesn't finish + pause = enqueue_store_result_job("pause", pause: 60.second) + + wait_while_with_timeout(1.second) { SolidQueue::ReadyExecution.count > 0 } + + signal_process(@pid, :QUIT, wait: 0.4.second) + wait_for_jobs_to_finish_for(2.seconds, except: pause) + + wait_while_with_timeout(2.seconds) { process_exists?(@pid) } + assert_not process_exists?(@pid) + + assert_completed_job_results("no pause") + assert_job_status(no_pause, :finished) + + assert_started_job_result("pause") + # Workers were shutdown without a chance to terminate orderly, but + # since they're linked to the supervisor, the supervisor deregistering + # also deregistered them and released claimed jobs + assert_job_status(pause, :ready) + assert_clean_termination + end + + private + def assert_clean_termination + wait_for_registered_processes 0, timeout: 0.2.second + assert_no_registered_processes + assert_no_claimed_jobs + assert_not process_exists?(@pid) + end + + def assert_registered_workers_for(*queues, supervisor_pid: nil) + workers = find_processes_registered_as("Worker") + registered_queues = workers.map { |process| process.metadata["queues"] }.compact + assert_equal queues.map(&:to_s).sort, registered_queues.sort + if supervisor_pid + assert_equal [ supervisor_pid ], workers.map { |process| process.supervisor.pid }.uniq + end + end + + def enqueue_store_result_job(value, queue_name = :background, **options) + StoreResultJob.set(queue: queue_name).perform_later(value, **options) + end + + def assert_completed_job_results(value, queue_name = :background, count = 1) + actual = skip_active_record_query_cache { + JobResult.where(queue_name: queue_name, status: "completed", value: value).count + } + assert_equal count, actual + end + + def assert_started_job_result(value, queue_name = :background, count = 1) + actual = skip_active_record_query_cache { + JobResult.where(queue_name: queue_name, status: "started", value: value).count + } + assert_equal count, actual + end + + def assert_job_status(active_job, status) + # Make sure we skip AR query cache. Otherwise the queries done here + # might be cached and since we haven't done any non-SELECT queries + # after they were cached on the connection used in the test, the cache + # will still apply, even though the data returned by the cached queries + # might have been deleted in the forked processes. + skip_active_record_query_cache do + job = SolidQueue::Job.find_by(active_job_id: active_job.job_id) + assert job.public_send("#{status}?") + end + end +end