https://github.com/mperham/sidekiq
Tip revision: a7038907bd8e114c4726ef1a2e4247294bd40cec authored by Mike Perham on 13 February 2023, 18:17:52 UTC
move perf lower
move perf lower
Tip revision: a703890
client.rb
# frozen_string_literal: true
require "securerandom"
require "sidekiq/middleware/chain"
require "sidekiq/job_util"
module Sidekiq
class Client
include Sidekiq::JobUtil
##
# Define client-side middleware:
#
# client = Sidekiq::Client.new
# client.middleware do |chain|
# chain.use MyClientMiddleware
# end
# client.push('class' => 'SomeJob', 'args' => [1,2,3])
#
# All client instances default to the globally-defined
# Sidekiq.client_middleware but you can change as necessary.
#
def middleware(&block)
if block
@chain = @chain.dup
yield @chain
end
@chain
end
attr_accessor :redis_pool
# Sidekiq::Client is responsible for pushing job payloads to Redis.
# Requires the :pool or :config keyword argument.
#
# Sidekiq::Client.new(pool: Sidekiq::RedisConnection.create)
#
# Inside the Sidekiq process, you can reuse the configured resources:
#
# Sidekiq::Client.new(config: config)
#
# @param pool [ConnectionPool] explicit Redis pool to use
# @param config [Sidekiq::Config] use the pool and middleware from the given Sidekiq container
# @param chain [Sidekiq::Middleware::Chain] use the given middleware chain
def initialize(*args, **kwargs)
if args.size == 1 && kwargs.size == 0
warn "Sidekiq::Client.new(pool) is deprecated, please use Sidekiq::Client.new(pool: pool), #{caller(0..3)}"
# old calling method, accept 1 pool argument
@redis_pool = args[0]
@chain = Sidekiq.default_configuration.client_middleware
@config = Sidekiq.default_configuration
else
# new calling method: keyword arguments
@config = kwargs[:config] || Sidekiq.default_configuration
@redis_pool = kwargs[:pool] || Thread.current[:sidekiq_redis_pool] || @config&.redis_pool
@chain = kwargs[:chain] || @config&.client_middleware
raise ArgumentError, "No Redis pool available for Sidekiq::Client" unless @redis_pool
end
end
##
# The main method used to push a job to Redis. Accepts a number of options:
#
# queue - the named queue to use, default 'default'
# class - the job class to call, required
# args - an array of simple arguments to the perform method, must be JSON-serializable
# at - timestamp to schedule the job (optional), must be Numeric (e.g. Time.now.to_f)
# retry - whether to retry this job if it fails, default true or an integer number of retries
# backtrace - whether to save any error backtrace, default false
#
# If class is set to the class name, the jobs' options will be based on Sidekiq's default
# job options. Otherwise, they will be based on the job class's options.
#
# Any options valid for a job class's sidekiq_options are also available here.
#
# All options must be strings, not symbols. NB: because we are serializing to JSON, all
# symbols in 'args' will be converted to strings. Note that +backtrace: true+ can take quite a bit of
# space in Redis; a large volume of failing jobs can start Redis swapping if you aren't careful.
#
# Returns a unique Job ID. If middleware stops the job, nil will be returned instead.
#
# Example:
# push('queue' => 'my_queue', 'class' => MyJob, 'args' => ['foo', 1, :bat => 'bar'])
#
def push(item)
normed = normalize_item(item)
payload = middleware.invoke(item["class"], normed, normed["queue"], @redis_pool) do
normed
end
if payload
verify_json(payload)
raw_push([payload])
payload["jid"]
end
end
##
# Push a large number of jobs to Redis. This method cuts out the redis
# network round trip latency. I wouldn't recommend pushing more than
# 1000 per call but YMMV based on network quality, size of job args, etc.
# A large number of jobs can cause a bit of Redis command processing latency.
#
# Takes the same arguments as #push except that args is expected to be
# an Array of Arrays. All other keys are duplicated for each job. Each job
# is run through the client middleware pipeline and each job gets its own Job ID
# as normal.
#
# Returns an array of the of pushed jobs' jids. The number of jobs pushed can be less
# than the number given if the middleware stopped processing for one or more jobs.
def push_bulk(items)
args = items["args"]
raise ArgumentError, "Bulk arguments must be an Array of Arrays: [[1], [2]]" unless args.is_a?(Array) && args.all?(Array)
return [] if args.empty? # no jobs to push
at = items.delete("at")
raise ArgumentError, "Job 'at' must be a Numeric or an Array of Numeric timestamps" if at && (Array(at).empty? || !Array(at).all? { |entry| entry.is_a?(Numeric) })
raise ArgumentError, "Job 'at' Array must have same size as 'args' Array" if at.is_a?(Array) && at.size != args.size
jid = items.delete("jid")
raise ArgumentError, "Explicitly passing 'jid' when pushing more than one job is not supported" if jid && args.size > 1
normed = normalize_item(items)
payloads = args.map.with_index { |job_args, index|
copy = normed.merge("args" => job_args, "jid" => SecureRandom.hex(12))
copy["at"] = (at.is_a?(Array) ? at[index] : at) if at
result = middleware.invoke(items["class"], copy, copy["queue"], @redis_pool) do
verify_json(copy)
copy
end
result || nil
}.compact
raw_push(payloads) unless payloads.empty?
payloads.collect { |payload| payload["jid"] }
end
# Allows sharding of jobs across any number of Redis instances. All jobs
# defined within the block will use the given Redis connection pool.
#
# pool = ConnectionPool.new { Redis.new }
# Sidekiq::Client.via(pool) do
# SomeJob.perform_async(1,2,3)
# SomeOtherJob.perform_async(1,2,3)
# end
#
# Generally this is only needed for very large Sidekiq installs processing
# thousands of jobs per second. I do not recommend sharding unless
# you cannot scale any other way (e.g. splitting your app into smaller apps).
def self.via(pool)
raise ArgumentError, "No pool given" if pool.nil?
current_sidekiq_pool = Thread.current[:sidekiq_redis_pool]
Thread.current[:sidekiq_redis_pool] = pool
yield
ensure
Thread.current[:sidekiq_redis_pool] = current_sidekiq_pool
end
class << self
def push(item)
new.push(item)
end
def push_bulk(items)
new.push_bulk(items)
end
# Resque compatibility helpers. Note all helpers
# should go through Sidekiq::Job#client_push.
#
# Example usage:
# Sidekiq::Client.enqueue(MyJob, 'foo', 1, :bat => 'bar')
#
# Messages are enqueued to the 'default' queue.
#
def enqueue(klass, *args)
klass.client_push("class" => klass, "args" => args)
end
# Example usage:
# Sidekiq::Client.enqueue_to(:queue_name, MyJob, 'foo', 1, :bat => 'bar')
#
def enqueue_to(queue, klass, *args)
klass.client_push("queue" => queue, "class" => klass, "args" => args)
end
# Example usage:
# Sidekiq::Client.enqueue_to_in(:queue_name, 3.minutes, MyJob, 'foo', 1, :bat => 'bar')
#
def enqueue_to_in(queue, interval, klass, *args)
int = interval.to_f
now = Time.now.to_f
ts = ((int < 1_000_000_000) ? now + int : int)
item = {"class" => klass, "args" => args, "at" => ts, "queue" => queue}
item.delete("at") if ts <= now
klass.client_push(item)
end
# Example usage:
# Sidekiq::Client.enqueue_in(3.minutes, MyJob, 'foo', 1, :bat => 'bar')
#
def enqueue_in(interval, klass, *args)
klass.perform_in(interval, *args)
end
end
private
def raw_push(payloads)
@redis_pool.with do |conn|
retryable = true
begin
conn.pipelined do |pipeline|
atomic_push(pipeline, payloads)
end
rescue RedisClient::Error => ex
# 2550 Failover can cause the server to become a replica, need
# to disconnect and reopen the socket to get back to the primary.
# 4495 Use the same logic if we have a "Not enough replicas" error from the primary
# 4985 Use the same logic when a blocking command is force-unblocked
# The retry logic is copied from sidekiq.rb
if retryable && ex.message =~ /READONLY|NOREPLICAS|UNBLOCKED/
conn.close
retryable = false
retry
end
raise
end
end
true
end
def atomic_push(conn, payloads)
if payloads.first.key?("at")
conn.zadd("schedule", payloads.flat_map { |hash|
at = hash.delete("at").to_s
[at, Sidekiq.dump_json(hash)]
})
else
queue = payloads.first["queue"]
now = Time.now.to_f
to_push = payloads.map { |entry|
entry["enqueued_at"] = now
Sidekiq.dump_json(entry)
}
conn.sadd("queues", [queue])
conn.lpush("queue:#{queue}", to_push)
end
end
end
end