NAME
Minion - Job queue
SYNOPSIS
use Minion;
# Connect to backend
my $minion = Minion->new(Pg => 'postgresql://postgres@/test');
# Add tasks
$minion->add_task(something_slow => sub {
my ($job, @args) = @_;
sleep 5;
say 'This is a background worker process.';
});
# Enqueue jobs
$minion->enqueue(something_slow => ['foo', 'bar']);
$minion->enqueue(something_slow => [1, 2, 3] => {priority => 5});
# Perform jobs for testing
$minion->enqueue(something_slow => ['foo', 'bar']);
$minion->perform_jobs;
# Start a worker to perform up to 12 jobs concurrently
my $worker = $minion->worker;
$worker->status->{jobs} = 12;
$worker->run;
DESCRIPTION
Minion is a high performance job queue for the Perl programming language, with support for multiple named queues, priorities, delayed jobs, job dependencies, job progress, job results, retries with backoff, rate limiting, unique jobs, expiring jobs, statistics, distributed workers, parallel processing, autoscaling, remote control, Mojolicious admin ui, resource leak protection and multiple backends (such as PostgreSQL).
Job queues allow you to process time and/or computationally intensive tasks in background processes, outside of the request/response lifecycle of web applications. Among those tasks you'll commonly find image resizing, spam filtering, HTTP downloads, building tarballs, warming caches and basically everything else you can imagine that's not super fast.
BASICS
You can use Minion as a standalone job queue or integrate it into Mojolicious applications with the plugin Mojolicious::Plugin::Minion.
use Mojolicious::Lite;
plugin Minion => {Pg => 'postgresql://sri:s3cret@localhost/test'};
# Slow task
app->minion->add_task(poke_mojo => sub {
my $job = shift;
$job->app->ua->get('mojolicious.org');
$job->app->log->debug('We have poked mojolicious.org for a visitor');
});
# Perform job in a background worker process
get '/' => sub {
my $c = shift;
$c->minion->enqueue('poke_mojo');
$c->render(text => 'We will poke mojolicious.org for you soon.');
};
app->start;
Background worker processes are usually started with the command Minion::Command::minion::worker, which becomes automatically available when an application loads Mojolicious::Plugin::Minion.
$ ./myapp.pl minion worker
The worker process will fork a new process for every job that is being processed. This allows for resources such as memory to be returned to the operating system once a job is finished. Perl fork is very fast, so don't worry about the overhead.
Minion::Worker
|- Minion::Job [1]
|- Minion::Job [2]
+- ...
By default up to four jobs will be processed in parallel, but that can be changed with configuration options or on demand with signals.
$ ./myapp.pl minion worker -j 12
Jobs can be managed right from the command line with Minion::Command::minion::job.
$ ./myapp.pl minion job
You can also add an admin ui to your application by loading the plugin Mojolicious::Plugin::Minion::Admin. Just make sure to secure access before making your application publically accessible.
# Make admin ui available under "/minion"
plugin 'Minion::Admin';
To manage background worker processes with systemd, you can use a unit configuration file like this.
[Unit]
Description=My Mojolicious application workers
After=postgresql.service
[Service]
Type=simple
ExecStart=/home/sri/myapp/myapp.pl minion worker -m production
KillMode=process
[Install]
WantedBy=multi-user.target
Every job can fail or succeed, but not get lost, the system is eventually consistent and will preserve job results for as long as you like, depending on "remove_after". While individual workers can fail in the middle of processing a job, the system will detect this and ensure that no job is left in an uncertain state, depending on "missing_after".
GROWING
And as your application grows, you can move tasks into application specific plugins.
package MyApp::Task::PokeMojo;
use Mojo::Base 'Mojolicious::Plugin';
sub register {
my ($self, $app) = @_;
$app->minion->add_task(poke_mojo => sub {
my $job = shift;
$job->app->ua->get('mojolicious.org');
$job->app->log->debug('We have poked mojolicious.org for a visitor');
});
}
1;
Which are loaded like any other plugin from your application.
# Mojolicious
$app->plugin('MyApp::Task::PokeMojo');
# Mojolicious::Lite
plugin 'MyApp::Task::PokeMojo';
TASK CLASSES
For even more flexibility you can also move tasks into dedicated classes. Allowing the use of Perl features such as inheritance and roles. But be aware that support for task classes is still EXPERIMENTAL and might change without warning!
package MyApp::Task::PokeMojo;
use Mojo::Base 'Minion::Job';
sub run {
my $self = shift;
$self->app->ua->get('mojolicious.org');
$self->app->log->debug('We have poked mojolicious.org for a visitor');
}
1;
Task classes are registered just like any other task with "add_task" and you can even register the same class with multiple names.
$minion->add_task(poke_mojo => 'MyApp::Task::PokeMojo');
EXAMPLES
This distribution also contains a great example application you can use for inspiration. The link checker will show you how to integrate background jobs into well-structured Mojolicious applications.
EVENTS
Minion inherits all events from Mojo::EventEmitter and can emit the following new ones.
enqueue
$minion->on(enqueue => sub {
my ($minion, $id) = @_;
...
});
Emitted after a job has been enqueued, in the process that enqueued it.
$minion->on(enqueue => sub {
my ($minion, $id) = @_;
say "Job $id has been enqueued.";
});
worker
$minion->on(worker => sub {
my ($minion, $worker) = @_;
...
});
Emitted in the worker process after it has been created.
$minion->on(worker => sub {
my ($minion, $worker) = @_;
say "Worker $$ started.";
});
ATTRIBUTES
Minion implements the following attributes.
app
my $app = $minion->app;
$minion = $minion->app(MyApp->new);
Application for job queue, defaults to a Mojo::HelloWorld object. Note that this attribute is weakened.
backend
my $backend = $minion->backend;
$minion = $minion->backend(Minion::Backend::Pg->new);
Backend, usually a Minion::Backend::Pg object.
backoff
my $cb = $minion->backoff;
$minion = $minion->backoff(sub {...});
A callback used to calculate the delay for automatically retried jobs, defaults to (retries ** 4) + 15
(15, 16, 31, 96, 271, 640...), which means that roughly 25
attempts can be made in 21
days.
$minion->backoff(sub {
my $retries = shift;
return ($retries ** 4) + 15 + int(rand 30);
});
missing_after
my $after = $minion->missing_after;
$minion = $minion->missing_after(172800);
Amount of time in seconds after which workers without a heartbeat will be considered missing and removed from the registry by "repair", defaults to 1800
(30 minutes).
remove_after
my $after = $minion->remove_after;
$minion = $minion->remove_after(86400);
Amount of time in seconds after which jobs that have reached the state finished
and have no unresolved dependencies will be removed automatically by "repair", defaults to 172800
(2 days). It is not recommended to set this value below 2 days.
stuck_after
my $after = $minion->stuck_after;
$minion = $minion->stuck_after(86400);
Amount of time in seconds after which jobs that have not been processed will be considered stuck by "repair" and transition to the failed
state, defaults to 172800
(2 days).
tasks
my $tasks = $minion->tasks;
$minion = $minion->tasks({foo => sub {...}});
Registered tasks.
METHODS
Minion inherits all methods from Mojo::EventEmitter and implements the following new ones.
add_task
$minion = $minion->add_task(foo => sub {...});
$minion = $minion->add_task(foo => 'MyApp::Task::Foo');
Register a task, which can be a closure or a custom Minion::Job subclass. Note that support for custom task classes is EXPERIMENTAL and might change without warning!
# Job with result
$minion->add_task(add => sub {
my ($job, $first, $second) = @_;
$job->finish($first + $second);
});
my $id = $minion->enqueue(add => [1, 1]);
my $result = $minion->job($id)->info->{result};
broadcast
my $bool = $minion->broadcast('some_command');
my $bool = $minion->broadcast('some_command', [@args]);
my $bool = $minion->broadcast('some_command', [@args], [$id1, $id2, $id3]);
Broadcast remote control command to one or more workers.
# Broadcast "stop" command to all workers to kill job 10025
$minion->broadcast('stop', [10025]);
# Broadcast "kill" command to all workers to interrupt job 10026
$minion->broadcast('kill', ['INT', 10026]);
# Broadcast "jobs" command to pause worker 23
$minion->broadcast('jobs', [0], [23]);
class_for_task
my $class = $minion->class_for_task('foo');
Return job class for task. Note that this method is EXPERIMENTAL and might change without warning!
enqueue
my $id = $minion->enqueue('foo');
my $id = $minion->enqueue(foo => [@args]);
my $id = $minion->enqueue(foo => [@args] => {priority => 1});
Enqueue a new job with inactive
state. Arguments get serialized by the "backend" (often with Mojo::JSON), so you shouldn't send objects and be careful with binary data, nested data structures with hash and array references are fine though.
These options are currently available:
- attempts
-
attempts => 25
Number of times performing this job will be attempted, with a delay based on "backoff" after the first attempt, defaults to
1
. - delay
-
delay => 10
Delay job for this many seconds (from now), defaults to
0
. - expire
-
expire => 300
Job is valid for this many seconds (from now) before it expires. Note that this option is EXPERIMENTAL and might change without warning!
- lax
-
lax => 1
Existing jobs this job depends on may also have transitioned to the
failed
state to allow for it to be processed, defaults tofalse
. Note that this option is EXPERIMENTAL and might change without warning! - notes
-
notes => {foo => 'bar', baz => [1, 2, 3]}
Hash reference with arbitrary metadata for this job that gets serialized by the "backend" (often with Mojo::JSON), so you shouldn't send objects and be careful with binary data, nested data structures with hash and array references are fine though.
- parents
-
parents => [$id1, $id2, $id3]
One or more existing jobs this job depends on, and that need to have transitioned to the state
finished
before it can be processed. - priority
-
priority => 5
Job priority, defaults to
0
. Jobs with a higher priority get performed first. - queue
-
queue => 'important'
Queue to put job in, defaults to
default
.
foreground
my $bool = $minion->foreground($id);
Retry job in minion_foreground
queue, then perform it right away with a temporary worker in this process, very useful for debugging.
guard
my $guard = $minion->guard('foo', 3600);
my $guard = $minion->guard('foo', 3600, {limit => 20});
Same as "lock", but returns a scope guard object that automatically releases the lock as soon as the object is destroyed, or undef
if aquiring the lock failed.
# Only one job should run at a time (unique job)
$minion->add_task(do_unique_stuff => sub {
my ($job, @args) = @_;
return $job->finish('Previous job is still active')
unless my $guard = $minion->guard('fragile_backend_service', 7200);
...
});
# Only five jobs should run at a time and we try again later if necessary
$minion->add_task(do_concurrent_stuff => sub {
my ($job, @args) = @_;
return $job->retry({delay => 30})
unless my $guard = $minion->guard('some_web_service', 60, {limit => 5});
...
});
history
my $history = $minion->history;
Get history information for job queue.
These fields are currently available:
- daily
-
daily => [{epoch => 12345, finished_jobs => 95, failed_jobs => 2}, ...]
Hourly counts for processed jobs from the past day.
is_locked
my $bool = $minion->is_locked('foo');
Check if a lock with that name is currently active.
job
my $job = $minion->job($id);
Get Minion::Job object without making any changes to the actual job or return undef
if job does not exist.
# Check job state
my $state = $minion->job($id)->info->{state};
# Get job metadata
my $progress = $minion->$job($id)->info->{notes}{progress};
# Get job result
my $result = $minion->job($id)->info->{result};
jobs
my $jobs = $minion->jobs;
my $jobs = $minion->jobs({states => ['inactive']});
Return Minion::Iterator object to safely iterate through job information.
# Iterate through jobs for two tasks
my $jobs = $minion->jobs({tasks => ['foo', 'bar']});
while (my $info = $jobs->next) {
say "$info->{id}: $info->{state}";
}
# Remove all failed jobs from a named queue
my $jobs = $minion->jobs({states => ['failed'], queues => ['unimportant']});
while (my $info = $jobs->next) {
$minion->job($info->{id})->remove;
}
# Count failed jobs for a task
say $minion->jobs({states => ['failed'], tasks => ['foo']})->total;
These options are currently available:
- ids
-
ids => ['23', '24']
List only jobs with these ids.
- notes
-
notes => ['foo', 'bar']
List only jobs with one of these notes.
- queues
-
queues => ['important', 'unimportant']
List only jobs in these queues.
- states
-
states => ['inactive', 'active']
List only jobs in these states.
- tasks
-
tasks => ['foo', 'bar']
List only jobs for these tasks.
These fields are currently available:
- args
-
args => ['foo', 'bar']
Job arguments.
- attempts
-
attempts => 25
Number of times performing this job will be attempted.
- children
-
children => ['10026', '10027', '10028']
Jobs depending on this job.
- created
-
created => 784111777
Epoch time job was created.
- delayed
-
delayed => 784111777
Epoch time job was delayed to.
- expires
-
expires => 784111777
Epoch time job is valid until before it expires.
- finished
-
finished => 784111777
Epoch time job was finished.
- id
-
id => 10025
Job id.
- lax
-
lax => 0
Existing jobs this job depends on may also have failed to allow for it to be processed.
- notes
-
notes => {foo => 'bar', baz => [1, 2, 3]}
Hash reference with arbitrary metadata for this job.
- parents
-
parents => ['10023', '10024', '10025']
Jobs this job depends on.
- priority
-
priority => 3
Job priority.
- queue
-
queue => 'important'
Queue name.
- result
-
result => 'All went well!'
Job result.
- retried
-
retried => 784111777
Epoch time job has been retried.
- retries
-
retries => 3
Number of times job has been retried.
- started
-
started => 784111777
Epoch time job was started.
- state
-
state => 'inactive'
Current job state, usually
active
,failed
,finished
orinactive
. - task
-
task => 'foo'
Task name.
- time
-
time => 78411177
Server time.
- worker
-
worker => '154'
Id of worker that is processing the job.
lock
my $bool = $minion->lock('foo', 3600);
my $bool = $minion->lock('foo', 3600, {limit => 20});
Try to acquire a named lock that will expire automatically after the given amount of time in seconds. You can release the lock manually with "unlock" to limit concurrency, or let it expire for rate limiting. For convenience you can also use "guard" to release the lock automatically, even if the job failed.
# Only one job should run at a time (unique job)
$minion->add_task(do_unique_stuff => sub {
my ($job, @args) = @_;
return $job->finish('Previous job is still active')
unless $minion->lock('fragile_backend_service', 7200);
...
$minion->unlock('fragile_backend_service');
});
# Only five jobs should run at a time and we wait for our turn
$minion->add_task(do_concurrent_stuff => sub {
my ($job, @args) = @_;
sleep 1 until $minion->lock('some_web_service', 60, {limit => 5});
...
$minion->unlock('some_web_service');
});
# Only a hundred jobs should run per hour and we try again later if necessary
$minion->add_task(do_rate_limited_stuff => sub {
my ($job, @args) = @_;
return $job->retry({delay => 3600})
unless $minion->lock('another_web_service', 3600, {limit => 100});
...
});
An expiration time of 0
can be used to check if a named lock could have been acquired without creating one.
# Check if the lock "foo" could have been acquired
say 'Lock could have been acquired' unless $minion->lock('foo', 0);
Or to simply check if a named lock already exists you can also use "is_locked".
These options are currently available:
- limit
-
limit => 20
Number of shared locks with the same name that can be active at the same time, defaults to
1
.
new
my $minion = Minion->new(Pg => 'postgresql://postgres@/test');
my $minion = Minion->new(Pg => Mojo::Pg->new);
Construct a new Minion object.
perform_jobs
$minion->perform_jobs;
$minion->perform_jobs({queues => ['important']});
Perform all jobs with a temporary worker, very useful for testing.
# Longer version
my $worker = $minion->worker;
while (my $job = $worker->register->dequeue(0)) { $job->perform }
$worker->unregister;
These options are currently available:
- queues
-
queues => ['important']
One or more queues to dequeue jobs from, defaults to
default
.
repair
$minion = $minion->repair;
Repair worker registry and job queue if necessary.
reset
$minion = $minion->reset({all => 1});
Reset job queue.
These options are currently available:
- all
-
all => 1
Reset everything.
- locks
-
locks => 1
Reset only locks.
result_p
my $promise = $minion->result_p($id);
my $promise = $minion->result_p($id, {interval => 5});
Return a Mojo::Promise object for the result of a job. The state finished
will result in the promise being fullfilled
, and the state failed
in the promise being rejected
. This operation can be cancelled by resolving the promise manually at any time.
# Enqueue job and receive the result at some point in the future
my $id = $minion->enqueue('foo');
$minion->result_p($id)->then(sub {
my $info = shift;
my $result = ref $info ? $info->{result} : 'Job already removed';
say "Finished: $result";
})->catch(sub {
my $info = shift;
say "Failed: $info->{result}";
})->wait;
These options are currently available:
- interval
-
interval => 5
Polling interval in seconds for checking if the state of the job has changed, defaults to
3
.
stats
my $stats = $minion->stats;
Get statistics for the job queue.
# Check idle workers
my $idle = $minion->stats->{inactive_workers};
These fields are currently available:
- active_jobs
-
active_jobs => 100
Number of jobs in
active
state. - active_locks
-
active_locks => 100
Number of active named locks.
- active_workers
-
active_workers => 100
Number of workers that are currently processing a job.
- delayed_jobs
-
delayed_jobs => 100
Number of jobs in
inactive
state that are scheduled to run at specific time in the future or have unresolved dependencies. - enqueued_jobs
-
enqueued_jobs => 100000
Rough estimate of how many jobs have ever been enqueued.
- failed_jobs
-
failed_jobs => 100
Number of jobs in
failed
state. - finished_jobs
-
finished_jobs => 100
Number of jobs in
finished
state. - inactive_jobs
-
inactive_jobs => 100
Number of jobs in
inactive
state. - inactive_workers
-
inactive_workers => 100
Number of workers that are currently not processing a job.
- uptime
-
uptime => 1000
Uptime in seconds.
unlock
my $bool = $minion->unlock('foo');
Release a named lock that has been previously acquired with "lock".
worker
my $worker = $minion->worker;
Build Minion::Worker object. Note that this method should only be used to implement custom workers.
# Use the standard worker with all its features
my $worker = $minion->worker;
$worker->status->{jobs} = 12;
$worker->status->{queues} = ['important'];
$worker->run;
# Perform one job manually in a separate process
my $worker = $minion->repair->worker->register;
my $job = $worker->dequeue(5);
$job->perform;
$worker->unregister;
# Perform one job manually in this process
my $worker = $minion->repair->worker->register;
my $job = $worker->dequeue(5);
if (my $err = $job->execute) { $job->fail($err) }
else { $job->finish }
$worker->unregister;
# Build a custom worker performing multiple jobs at the same time
my %jobs;
my $worker = $minion->repair->worker->register;
do {
for my $id (keys %jobs) {
delete $jobs{$id} if $jobs{$id}->is_finished;
}
if (keys %jobs >= 4) { sleep 5 }
else {
my $job = $worker->dequeue(5);
$jobs{$job->id} = $job->start if $job;
}
} while keys %jobs;
$worker->unregister;
workers
my $workers = $minion->workers;
my $workers = $minion->workers({ids => [2, 3]});
Return Minion::Iterator object to safely iterate through worker information.
# Iterate through workers
my $workers = $minion->workers;
while (my $info = $workers->next) {
say "$info->{id}: $info->{host}";
}
These options are currently available:
- ids
-
ids => ['23', '24']
List only workers with these ids.
These fields are currently available:
- id
-
id => 22
Worker id.
- host
-
host => 'localhost'
Worker host.
- jobs
-
jobs => ['10023', '10024', '10025', '10029']
Ids of jobs the worker is currently processing.
- notified
-
notified => 784111777
Epoch time worker sent the last heartbeat.
- pid
-
pid => 12345
Process id of worker.
- started
-
started => 784111777
Epoch time worker was started.
- status
-
status => {queues => ['default', 'important']}
Hash reference with whatever status information the worker would like to share.
API
This is the class hierarchy of the Minion distribution.
BUNDLED FILES
The Minion distribution includes a few files with different licenses that have been bundled for internal use.
Minion Artwork
Copyright (C) 2017, Sebastian Riedel.
Licensed under the CC-SA License, Version 4.0 http://creativecommons.org/licenses/by-sa/4.0.
Bootstrap
Copyright (C) 2011-2018 The Bootstrap Authors.
Licensed under the MIT License, http://creativecommons.org/licenses/MIT.
D3.js
Copyright (C) 2010-2016, Michael Bostock.
Licensed under the 3-Clause BSD License, https://opensource.org/licenses/BSD-3-Clause.
epoch.js
Copyright (C) 2014 Fastly, Inc.
Licensed under the MIT License, http://creativecommons.org/licenses/MIT.
Font Awesome
Copyright (C) Dave Gandy.
Licensed under the MIT License, http://creativecommons.org/licenses/MIT, and the SIL OFL 1.1, http://scripts.sil.org/OFL.
moment.js
Copyright (C) JS Foundation and other contributors.
Licensed under the MIT License, http://creativecommons.org/licenses/MIT.
popper.js
Copyright (C) Federico Zivolo 2017.
Licensed under the MIT License, http://creativecommons.org/licenses/MIT.
AUTHOR
Sebastian Riedel, sri@cpan.org
.
CREDITS
In alphabetical order:
Andrey Khozov
Andrii Nikitin
Brian Medley
Franz Skale
Hubert "depesz" Lubaczewski
Joel Berger
Paul Williams
Stefan Adams
COPYRIGHT AND LICENSE
Copyright (C) 2014-2020, Sebastian Riedel and others.
This program is free software, you can redistribute it and/or modify it under the terms of the Artistic License version 2.0.
SEE ALSO
https://github.com/mojolicious/minion, https://minion.pm, Mojolicious::Guides, https://mojolicious.org.