Class: Sidekiq::JobRetry

Inherits:
Object
  • Object
show all
Includes:
Util
Defined in:
lib/sidekiq/job_retry.rb

Overview

Automatically retry jobs that fail in Sidekiq. Sidekiq's retry support assumes a typical development lifecycle:

0. Push some code changes with a bug in it.
1. Bug causes job processing to fail, Sidekiq's middleware captures
   the job and pushes it onto a retry queue.
2. Sidekiq retries jobs in the retry queue multiple times with
   an exponential delay, the job continues to fail.
3. After a few days, a developer deploys a fix. The job is
   reprocessed successfully.
4. Once retries are exhausted, Sidekiq will give up and move the
   job to the Dead Job Queue (aka morgue) where it must be dealt with
   manually in the Web UI.
5. After 6 months on the DJQ, Sidekiq will discard the job.

A job looks like:

{ 'class' => 'HardWorker', 'args' => [1, 2, 'foo'], 'retry' => true }

The 'retry' option also accepts a number (in place of 'true'):

{ 'class' => 'HardWorker', 'args' => [1, 2, 'foo'], 'retry' => 5 }

The job will be retried this number of times before giving up. (If simply 'true', Sidekiq retries 25 times)

We'll add a bit more data to the job to support retries:

* 'queue' - the queue to use
* 'retry_count' - number of times we've retried so far.
* 'error_message' - the message from the exception
* 'error_class' - the exception class
* 'failed_at' - the first time it failed
* 'retried_at' - the last time it was retried
* 'backtrace' - the number of lines of error backtrace to store

We don't store the backtrace by default as that can add a lot of overhead to the job and everyone is using an error service, right?

The default number of retries is 25 which works out to about 3 weeks You can change the default maximum number of retries in your initializer:

Sidekiq.options[:max_retries] = 7

or limit the number of retries for a particular worker with:

class MyWorker
  include Sidekiq::Worker
  sidekiq_options :retry => 10
end

Defined Under Namespace

Classes: Skip

Constant Summary collapse

DEFAULT_MAX_RETRY_ATTEMPTS =
25

Constants included from Util

Util::EXPIRY

Instance Method Summary collapse

Methods included from Util

#fire_event, #hostname, #identity, #logger, #process_nonce, #redis, #safe_thread, #watchdog

Methods included from ExceptionHandler

#handle_exception

Constructor Details

#initialize(options = {}) ⇒ JobRetry

Returns a new instance of JobRetry.


65
66
67
# File 'lib/sidekiq/job_retry.rb', line 65

def initialize(options = {})
  @max_retries = Sidekiq.options.merge(options).fetch(:max_retries, DEFAULT_MAX_RETRY_ATTEMPTS)
end

Instance Method Details

#global(msg, queue) ⇒ Object

The global retry handler requires only the barest of data. We want to be able to retry as much as possible so we don't require the worker to be instantiated.


72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
# File 'lib/sidekiq/job_retry.rb', line 72

def global(msg, queue)
  yield
rescue Skip => ex
  raise ex
rescue Sidekiq::Shutdown => ey
  # ignore, will be pushed back onto queue during hard_shutdown
  raise ey
rescue Exception => e
  # ignore, will be pushed back onto queue during hard_shutdown
  raise Sidekiq::Shutdown if exception_caused_by_shutdown?(e)

  raise e unless msg['retry']
  attempt_retry(nil, msg, queue, e)
  raise e
end

#local(worker, msg, queue) ⇒ Object

The local retry support means that any errors that occur within this block can be associated with the given worker instance. This is required to support the `sidekiq_retries_exhausted` block.

Note that any exception from the block is wrapped in the Skip exception so the global block does not reprocess the error. The Skip exception is unwrapped within Sidekiq::Processor#process before calling the handle_exception handlers.


97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
# File 'lib/sidekiq/job_retry.rb', line 97

def local(worker, msg, queue)
  yield
rescue Skip => ex
  raise ex
rescue Sidekiq::Shutdown => ey
  # ignore, will be pushed back onto queue during hard_shutdown
  raise ey
rescue Exception => e
  # ignore, will be pushed back onto queue during hard_shutdown
  raise Sidekiq::Shutdown if exception_caused_by_shutdown?(e)

  if msg['retry'] == nil
    msg['retry'] = worker.class.get_sidekiq_options['retry']
  end

  raise e unless msg['retry']
  attempt_retry(worker, msg, queue, e)
  # We've handled this error associated with this job, don't
  # need to handle it at the global level
  raise Skip
end