Bounce Processor Concepts

• Table of Contents
• Overview
• SimpleMH Requirements
• Raw Injection Requirements
  • Set the ListID
  • Set the Return-Path
• Events Triggered
• Types of Bounces (Hard/Soft/Other)
• Bounce Codes
• Repeat Bounce Counting System
  • Examples
  • Considerations for Hard Bounces
• Viewing Data
• Synchronous and Asynchronous Bounce Messages
  • Synchronous Bounce Messages
  • Asynchronous Bounce Messages
  • Considerations
• Service and Logfile
• Bounce Processor Queue

Overview

This page describes GreenArrow’s bounce processor from a conceptual point of view. See the Bounce Processor Configuration page for configuration instructions.

The bounce processor is responsible for processing bounce messages received at Bounce Mailboxes. The bounces are parsed to determine the reason for the bounce, along with the recipient address, ListID, and the SendID of the message. Each processed bounce triggers a bounce_all event.

Bounces are attributed to the email address that GreenArrow sent the email to. This means that if GreenArrow sends an email to AddressA, which forwards to AddressB, and AddressB generates a bounce message, GreenArrow will interpret the bounce as being for AddressA.

The bounce processor tracks the number of bounces to an email address, and the amount of time the email address has been bouncing. It combines this information with its configuration to determine if the email address is no longer valid. If the email address is no longer valid, then the bounce processor issues a bounce_bad_address event so you can deactivate the email address in your database.

When bounce_all and bounce_bad_address events occur, the Event Notification System can notify you via an SQL Query, HTTP Post, or one of its other event delivery methods.

The bounce processor works hard to avoid deactivating valid addresses, as well as to make sure that all invalid addresses are identified.

Proper bounce processing is essential for deliverability, as a sender can be blocked for sending to a high percentage of invalid addresses.

SimpleMH Requirements

Unlike with Raw Injection (described below), the ListID and Return-Path settings are not required for messages injected using SimpleMH. SimpleMH automatically sets the ListID and Return-Path based on the configuration of the Mail Class.

Because some systems incorrectly limit the localpart of an email address to 64 characters/bytes, we recommend that you keep the sum of the following to less than 47 characters/bytes (the extra 17 characters/bytes are reserved for GreenArrow metadata/separators):

• the length of the ListID configured on the Mail Class
• the length of the Mail Class name
• the length of the InstanceID (or 6 for the YYMMDD format if you do not specify an InstanceID)
• the length of the localpart of the bounce mailbox

Raw Injection Requirements

If you are using Raw Injection, you must configure a couple of things in the messages you send.

Set the ListID

You must set the ListID to a correct and consistent value. The repeat bounce counting is done on a per-ListID basis, so the ListID must be consistent. The ListID is compared case-insensitively for repeat bounce counting.

The ListID is described on the GreenArrow Concepts, and the X-GreenArrow-ListID header, which is used to set the ListID is described on the Raw Injection Method Headers page.

Set the Return-Path

The Return-Path is the field in an email message that determines where bounces are sent. You must correctly set the Return-Path on outgoing messages for the bounce processor to work.

The Return-Path should be set to a valid Bounce Mailbox’s address.

Details of how to set the Return-Path are in the Raw Injection Method Headers page.

Events Triggered

There are two event types triggered by the bounce processor:

1. bounce_all -- Triggered for every bounce received. (Do not use this to deactivate email addresses in your database.)
2. bounce_bad_address -- Triggered when an address is detected as bad and should be deactivated in your database.

Types of Bounces (Hard/Soft/Other)

GreenArrow breaks bounce messages down into three different types:

Bounce Codes

Bounce messages are categorized into different bounce codes. Each code has a default type (Hard, Soft or Other).

Repeat Bounce Counting System

With soft bounces, we want to remove the email address if the email address continues to soft bounce over a period of time. GreenArrow Engine’s repeat bounce counting system implements this.

You can configure the how many times an email address / ListID combination must soft bounce before it is marked as bad, and therefore removed from your list. You can also configure similar requirements for hard bounces. (See note below on setting this for hard bounces.)

The system is designed so that it does the right thing with the typical settings regardless of your mailing frequency -- if you are sending daily, weekly, monthly, or only occasionally.

These requirements must be met for an email address to be marked as bad:

1. The required number of soft or hard bounces must be received for the email address and ListID combination. (This is the hard_req_count and soft_req_count settings.)
2. All emails to that email address and ListID combination must have bounced, except that a configurable number may be allowed not to have bounced, (which is the hard_allowed_missing and soft_allowed_missing settings).
3. The first bounce and the last bounce must be at least a certain number of days apart. (This is the hard_req_days and soft_req_days settings.)

Rule #1 is required because we want to make sure that the email address has bounced a certain number of times before we mark it as bad.

Rule #2 is required because we want to make sure that the email address is consistently bouncing, not just occasionally bouncing. For example, if the only requirement was that the email address bounce 10 times, and if we sent daily emails but the email address bounced once per week, without this rule the address would get removed after 10 weeks (because it would have reached 10 bounces). With the hard_allow_missing and soft_allow_missing parameters, you can allow a few messages to be sent without producing bounces. This can be helpful if you require a large number of bounces because email is not perfect and sometimes messages get lost without producing bounces.

Rule #3 is required because we want to make sure that the email address is consistently bouncing over a long enough period to be sure it is bad. For example, if you send daily mailings, and if a user goes over quota for one week, you don’t want the 7 bounces generated to trigger a removal. By having soft_req_days set to 14, you would require that the error condition exists for at least two weeks.

Regardless of whether or not stateless event tracking is configured, repeat bounce counting only takes into account the bounces for the particular server on which they are processed.

Examples

With these settings:

Example #1

This bounce history would not cause the email to get removed because we have four bounces (which qualifies) but only four days between the first and last bounce (which does not qualify):

• Email sent Jan 1 – bounced
• Email sent Jan 3 – bounced
• Email sent Jan 4 – bounced
• Email sent Jan 5 – bounced

Example #2

This bounce history would not cause the email to get removed because we have only two bounces (which does not qualify) and eight days between the first and last bounce (which does qualify):

• Email sent Jan 1 – bounced
• Email sent Jan 9 – bounced

Example #3

This bounce history would cause the email to get removed because we have four bounces (which qualifies) and 8 days between the first and last bounce (which qualifies):

• Email sent Jan 1 – bounced
• Email sent Jan 3 – bounced
• Email sent Jan 6 – bounced
• Email sent Jan 9 – bounced

Example #4

This bounce history would cause the email to get removed, because we have four bounces (which qualifies) and 8 days between the first and last bounce (which qualifies), and only one missing bounce (the email sent on Jan 7) (which qualifies):

• Email sent Jan 1 – bounced
• Email sent Jan 3 – bounced
• Email sent Jan 6 – bounced
• Email sent Jan 7 – did not bounce
• Email sent Jan 9 – bounced

Example #5

This bounce history would not cause the email to get removed, because we have four bounces (which qualifies) and 8 days between the first and last bounce (which qualifies), but two missing bounces (the email sent on Jan 5 and Jan 7) (which does not qualify):

• Email sent Jan 1 – bounced
• Email sent Jan 3 – bounced
• Email sent Jan 5 – did not bounce
• Email sent Jan 6 – bounced
• Email sent Jan 7 – did not bounce
• Email sent Jan 9 – bounced

Considerations for Hard Bounces

GreenArrow allows you to set up repeat bounce counting for hard bounces, but we recommend being very careful with this.

The reason you may want to require more than one hard bounce is that very rarely an ISP will send a hard bounce for an email address that actually exists. Often this is a software error at the ISP. It sounds strange, but we have seen this happen.

The caution is that if you are starting to send to a list that has not had correct bounce processing done (perhaps imported from other software that was not processing bounces correctly), it’s imperative for deliverability that you get the hard bouncing addresses removed as quickly as possible, so you want to deactivate addresses after a single hard bounce.

If your list is already clean, and your hard bounce percentage is low, then to protect against this you can set hard_req_count=2 and hard_req_days=2.

Viewing Data

You can view statistical information about bounces by looking at a Send’s Statistics in GreenArrow Engine’s web interface. For a particular send, click the Bounce stats tab to view these stats.

Synchronous and Asynchronous Bounce Messages

There are two kinds of bounce messages, depending on when the failure happens:

Synchronous Bounce Messages

This is where a remote SMTP server rejects the message in the SMTP conversation with a 5xx error code. (Or if the message is continually deferred, once it reaches the maximum retry time, GreenArrow Engine will consider this an SMTP failure.) Internally, GreenArrow creates a bounce message and sends this to the bounce processor. This bounce message is synchronous.

“Synchronous” means same time -- these are bounce messages were the failure happens at the same time as the SMTP conversation.

Asynchronous Bounce Messages

This is where a remote SMTP server accepts the message via SMTP, and then later determines that it can not deliver the email, and returns a bounce message back to the Return-Path address of the message. This bounce message is an asynchronous bounce message.

This message will show accepted via SMTP, but then later a bounce will be logged.

“Asynchronous” means NOT at the same time -- these are bounce messages were the failure happens NOT at the same time as the SMTP conversation.

Considerations

The GreenArrow bounce processor processes both synchronous and asynchronous bounce messages.

In the reporting, synchronous bounces contribute to (a) the SMTP failures in the GreenArrow stats (the SMTP Server Stats tab, the TOTAL row, and the Failure column), and (b) the total number of bounces (the Bounce stats tab, the Total Bounces number. Asynchronous bounces only contribute to the total number of bounces (b). This is because asynchronous bounces don’t have a failure in the SMTP conversation.

Service and Logfile

Bounces and spam complaints are processed and logged by the same service. To monitor this service’s log file, execute the following command:

tail -F /var/hvmail/log/bounce-processor/current | tai64nlocal

The bounce processor is run from the /service/hvmail-bounce-processor service.

To temporarily turn off the bounce processor, run:

svc -d /service/hvmail-bounce-processor

To turn it back on run:

svc -u /service/hvmail-bounce-processor

To check whether the service is on, run:

hvmail_init status | grep hvmail-bounce-processor

Bounce Processor Queue

As bounces occur, they get delivered to GreenArrow’s bounce processing queue.

Normally GreenArrow’s bounce processor can keep up with the rate at which bounces are queued. You can verify that it is by running the hvmail_status status command, then looking at the “Bounce processor queue” section of the output. For example, the following output shows that there are 0 messages in the bounce processor’s queue:

Bounce processor queue:           0% used (     0/  3000) soft limit
                                  0% used (     0/  5000) hard limit

The “hard” and “soft” limits referenced in the above output are described in the Bounce Processor Configuration page.