On this page
In this guide, you can learn how to handle errors in your sink connector. The following list shows some common scenarios that cause your sink connector to experience an error:
You write to a topic using Avro serialization and try to decode your messages from that topic using Protobuf deserialization
You use a change data capture handler on a message that does not contain change event documents
You apply an invalid single message transform to incoming documents
When your sink connector encounters an error it does two actions:
When your connector encounters an error, it needs to handle it in some way. Your sink connector can do the following in response to an error:
By default, your sink connector terminates and stops processing messages when it encounters an error. This is a good option for you if any error in your sink connector indicates a serious problem.
When your sink connector crashes, you must do one of the following actions and then restart your connector to resume processing messages:
Allow your sink connector to temporarily tolerate errors
Update your sink connector's configuration to allow it to process the message
Remove the errant message from your topic
You can have your sink connector stop when it encounters an error by either not
specifying any value for the
errors.tolerance option, or by
adding the following to your connector configuration:
You can configure your sink connector to tolerate all errors and never stop processing messages. This is a good option for getting your sink connector up and running quickly, but you run the risk of missing problems in your connector as you do not receive any feedback if something goes wrong.
You can have your sink connector tolerate all errors by specifying the following option:
You can configure your sink connector to write errors and errant messages to a topic, called a dead letter queue, for you to inspect or process further. A dead letter queue is a location in message queueing systems such as Apache Kafka where the system routes errant messages instead of crashing or ignoring the error. Dead letter queues combine the feedback of stopping the program with the durability of tolerating all errors, and are a good error handling starting point for most deployments.
You can have your sink connector route all errant messages to a dead letter queue by specifying the following options:
errors.tolerance=all errors.deadletterqueue.topic.name=<name of topic to use as dead letter queue>
If you want to include the specific reason for the error as well as the errant message, use the following option:
Bulk Write Errors
If an error occurs while performing a bulk write to MongoDB, you must specify the following to log any information about the error to a dead letter queue:
errors.tolerance=all mongo.errors.tolerance=none errors.deadletterqueue.topic.name=<name of topic to use as dead letter queue>
We will fix this special case soon. To track the status of this issue, see this JIRA ticket.
For more information on the
mongo.errors.tolerance option, see our
section on connector level options.
For more information, see Confluent's guide on Dead Letter Queues.
To view another dead letter queue configuration example, see Dead Letter Queue Configuration Example.
You can record tolerated and untolerated errors to a log file. Click on the tabs to see how to log errors:
If you would like to log metadata about your message, such as your message's topic and offset, use the following option:
For more information, see Confluent's guide on logging with Kafka Connect.
The MongoDB Kafka Connector provides options that allow you to configure error handling at the connector level. The options are as follows:
Kafka Connect Option
MongoDB Kafka Connector Option
You want to use these options if you want your connector to respond differently to errors related to MongoDB than to errors related to the Kafka Connect framework.
For more information, see the following resources: