Retry

RetryProcessor

RetryProcessor is a swarrot processor. Its goal is to re-published messages in broker when an error occurred.

Configuration

Key	Default	Description
retry_key_pattern		The pattern to use to construct routing key (ie: key_%attempt%)
retry_key_generator		The callable to construct routing key (ie: (int $attempt, Message $message): string)
retry_attempts	3	The number of attempts before raising an exception.
retry_log_levels_map	[]	Map of classes to a log level when retry. (Warning by default)
retry_fail_log_levels_map	[]	Map of classes to a log level when all retries failed. (Warning by default)

Configuring either retry_key_pattern or retry_key_generator is mandatory.

How it works

When your processor throw an exception (= failed to process a message), the Retry Processor will catch it, and republish the message in a given exchange with the retry_key_pattern.

For example if the given MessagePublisherInterface is configured to publish in the exchange named retry and if the retry_key_pattern is key_%attempts%, when an exception is thrown, the RetryProcessor will publish a new message (similar to the first one) in the exchange retry with the routing_key key_1.

Real example

Let's say you want to consume a queue named mail. You would like to have 3 retries if the mail is not sent correctly. The first retry 30 seconds after, the second 3 minutes and the last one half an hour.

First, let's create the simple mail workflow.

• Create an exchange mail (type: direct)
• Create an queue mail
• Bind the exchange mail to the queue mail with routing_key mail

Now, to handle errors, let's say we republish all errors to an exchange retry. For our mail, the retry_key_pattern will be mail_retry_%attempts%. Here is what we need to have the first retry working:

• Create an exchange retry (type: direct)
• Create a queue mail_retry_1 with following arguments: x-message-ttl: 30000 (our 30 seconds), x-dead-letter-exchange: mail (the first exchange we created), x-dead-letter-routing-key: mail
• Bind the exchange retry to the queue mail_retry_1 with routing_key mail_retry_1

Once this configuration is done, when an exception is thrown, the RetryProcessor will publish a new message (similar to the first one) to the exchange retry with the routing_key mail_retry_1. This message will be route to the queue mail_retry_1. As there is a ttl, 30 seconds after its arrival, the message will be sent to the configured dead-letter exchange with the given dead-letter routing-key. In our case the message will go into the exchange mail with the routing key mail. Due to our first configuration, it will finish in the queue mail which is consumed by our consumer.

Of course if we want more than only one retry, we can create new queues mail_retry_X with the same configuration. The RetryProcessor will use the swarrot_retry_attempts key in the message headers to determine if the message must be retried or not.