Skip to content

kumo.on('requeue_message', function(message))

Since: Dev Builds Only

The functionality described in this section requires a dev build of KumoMTA. You can obtain a dev build by following the instructions in the Installation section.

This event is triggered when a message encountered a transient failure. Its purpose is to allow you to re-bind the message to an alternative queue, perhaps to relay it via an alternate tier or to use an alternative pool for egress.

Multiple instances of the requeue_message event can be registered, and they will be called in the order in which they were registered, until all registered events are called, or until one explicitly returns nil to signal that no more should be triggered.

The event is triggered prior to incrementing the number of attempts, so message:num_attempts() will return one less than the current number.

In order to re-bind the message you will typically alter one or more of the meta values of the message that impact the queue name:

  • queue
  • routing_domain
  • tenant
  • campaign

See Queues for more information.

If the effective queue name for the message is changed as a result of dispatching the message_requeued event, then the message will be immediately eligible for delivery in the context of its new queue, however, if the message has scheduling constraints set via msg:set_scheduling those will remain in effect unless you explicitly clear them. The reason for this is that kumod doesn't have any implicit knowledge of the semantics of the queue, so it doesn't know whether the scheduling constraints should remain in force or not.

In the example below, a message is re-routed to a smart host after the third attempt to send it encounters a transient failure.

local SMART_HOST = '[]'

kumo.on('message_requeued', function(msg)
  local queue = msg:get_queue_name()
  if queue ~= SMART_HOST and msg:num_attempts() >= 2 then
    -- Re-route to alternative infra to manage the rest of the send
    msg:set_meta('queue', SMART_HOST)
    -- clear any scheduling constraints, as they do not apply
    -- when sending via a smart host