kafka-connect-mq-source is a Kafka Connect source connector for copying data from IBM MQ into Apache Kafka.
The connector is supplied as source code which you can easily build into a JAR file.
Note: A sink connector for IBM MQ is also available on GitHub.
- Building the connector
- Running the connector
- Running the connector with Docker
- Deploying the connector to Kubernetes
- Data formats
- Security
- Configuration
- Exactly-once message delivery semantics
- Troubleshooting
- Support
- Issues and contributions
- License
To build the connector, you must have the following installed:
- git
- Maven 3.0 or later
- Java 8 or later
Clone the repository with the following command:
git clone https://github.com/ibm-messaging/kafka-connect-mq-source.gitChange directory into the kafka-connect-mq-source directory:
cd kafka-connect-mq-sourceBuild the connector using Maven:
mvn clean packageOnce built, the output is a single JAR called target/kafka-connect-mq-source-<version>-jar-with-dependencies.jar which contains all of the required dependencies.
NOTE: With the 2.0.0 release the base Kafka Connect library has been updated from 2.6.0 to 3.4.0 to enable the implementation of the exactly-once delivery.
For step-by-step instructions, see the following guides for running the connector:
- connecting to Apache Kafka running locally
- connecting to an installation of IBM Event Streams
To run the connector, you must have:
- The JAR from building the connector
- A properties file containing the configuration for the connector
- Apache Kafka 2.0.0 or later, either standalone or included as part of an offering such as IBM Event Streams (Apache Kafka 3.3.0 or later is required for exactly-once delivery).
- IBM MQ v8 or later, or the IBM MQ on Cloud service
The connector can be run in a Kafka Connect worker in either standalone (single process) or distributed mode. It's a good idea to start in standalone mode.
You need two configuration files, one for the configuration that applies to all of the connectors such as the Kafka bootstrap servers, and another for the configuration specific to the MQ source connector such as the connection information for your queue manager. For the former, the Kafka distribution includes a file called connect-standalone.properties that you can use as a starting point. For the latter, you can use config/mq-source.properties in this repository.
The connector connects to MQ using either a client or a bindings connection. For a client connection, you must provide the name of the queue manager, the connection name (one or more host/port pairs) and the channel name. In addition, you can provide a user name and password if the queue manager is configured to require them for client connections. If you look at the supplied config/mq-source.properties, you'll see how to specify the configuration required. For a bindings connection, you must provide provide the name of the queue manager and also run the Kafka Connect worker on the same system as the queue manager.
To run the connector in standalone mode from the directory into which you installed Apache Kafka, you use a command like this:
bin/connect-standalone.sh connect-standalone.properties mq-source.propertiesYou need an instance of Kafka Connect running in distributed mode. The Kafka distribution includes a file called connect-distributed.properties that you can use as a starting point, or follow Running with Docker or Deploying to Kubernetes
To start the MQ connector, you can use config/mq-source.json in this repository after replacing all placeholders and use a command like this:
curl -X POST -H "Content-Type: application/json" http://localhost:8083/connectors \
--data "@./config/mq-source.json"This repository includes an example Dockerfile to run Kafka Connect in distributed mode. It also adds in the MQ source connector as an available connector plugin. It uses the default connect-distributed.properties and connect-log4j.properties files.
mvn clean packagedocker build -t kafkaconnect-with-mq-source:2.0.0 .docker run -p 8083:8083 kafkaconnect-with-mq-source:2.0.0
NOTE: To provide custom properties files create a folder called config containing the connect-distributed.properties and connect-log4j.properties files and use a Docker volume to make them available when running the container like this:
docker run -v $(pwd)/config:/opt/kafka/config -p 8083:8083 kafkaconnect-with-mq-source:2.0.0To start the MQ connector, you can use config/mq-source.json in this repository after replacing all placeholders and use a command like this:
curl -X POST -H "Content-Type: application/json" http://localhost:8083/connectors \
--data "@./config/mq-source.json"This repository includes a Kubernetes yaml file called kafka-connect.yaml. This will create a deployment to run Kafka Connect in distributed mode and a service to access the deployment.
The deployment assumes the existence of a Secret called connect-distributed-config and a ConfigMap called connect-log4j-config. These can be created using the default files in your Kafka install, however it is easier to edit them later if comments and whitespaces are trimmed before creation.
Create Secret for Kafka Connect configuration:
cp kafka/config/connect-distributed.properties connect-distributed.properties.origsed '/^#/d;/^[[:space:]]*$/d' < connect-distributed.properties.orig > connect-distributed.propertieskubectl -n <namespace> create secret generic connect-distributed-config --from-file=connect-distributed.properties
Create ConfigMap for Kafka Connect Log4j configuration:
cp kafka/config/connect-log4j.properties connect-log4j.properties.origsed '/^#/d;/^[[:space:]]*$/d' < connect-log4j.properties.orig > connect-log4j.propertieskubectl -n <namespace> create configmap connect-log4j-config --from-file=connect-log4j.properties
NOTE: You will need to build the Docker image and push it to your Kubernetes image repository. Remember that the supplied Dockerfile is just an example and you will have to modify it for your needs. You might need to update the image name in the kafka-connect.yaml file.
- Update the namespace in
kafka-connect.yaml kubectl -n <namespace> apply -f kafka-connect.yamlcurl <serviceIP>:<servicePort>/connector-pluginsto see whether the MQ source connector is available to use
This repository includes a Kubernetes yaml file called strimzi.kafkaconnector.yaml for use with the Strimzi operator. Strimzi provides a simplified way of running the Kafka Connect distributed worker, by defining either a KafkaConnect resource or a KafkaConnectS2I resource.
The KafkaConnectS2I resource provides a nice way to have OpenShift do all the work of building the Docker images for you. This works particularly nicely combined with the KafkaConnector resource that represents an individual connector.
The following instructions assume you are running on OpenShift and have Strimzi 0.16 or later installed.
- Create a file called
kafka-connect-s2i.yamlcontaining the definition of a KafkaConnectS2I resource. You can use the examples in the Strimzi project to get started. - Configure it with the information it needs to connect to your Kafka cluster. You must include the annotation
strimzi.io/use-connector-resources: "true"to configure it to use KafkaConnector resources so you can avoid needing to call the Kafka Connect REST API directly. oc apply -f kafka-connect-s2i.yamlto create the cluster, which usually takes several minutes.
mvn clean packageto build the connector JAR.mkdir my-pluginscp target/kafka-connect-mq-source-*-jar-with-dependencies.jar my-pluginsoc start-build <kafkaconnectClusterName>-connect --from-dir ./my-pluginsto add the MQ source connector to the Kafka Connect distributed worker cluster. Wait for the build to complete, which usually takes a few minutes.oc describe kafkaconnects2i <kafkaConnectClusterName>to check that the MQ source connector is in the list of available connector plugins.
cp deploy/strimzi.kafkaconnector.yaml kafkaconnector.yaml- Update the
kafkaconnector.yamlfile to replace all of the values in<>, adding any additional configuration properties. oc apply -f kafkaconnector.yamlto start the connector.oc get kafkaconnectorto list the connectors. You can useoc describeto get more details on the connector, such as its status.
Kafka Connect is very flexible but it's important to understand the way that it processes messages to end up with a reliable system. When the connector encounters a message that it cannot process, it stops rather than throwing the message away. Therefore, you need to make sure that the configuration you use can handle the messages the connector will process.
This is rather complicated and it's likely that a future update of the connector will simplify matters.
Each message in Kafka Connect is associated with a representation of the message format known as a schema. Each Kafka message actually has two parts, key and value, and each part has its own schema. The MQ source connector does not currently make much use of message keys, but some of the configuration options use the word Value because they refer to the Kafka message value.
When the MQ source connector reads a message from MQ, it chooses a schema to represent the message format and creates an internal object called a record containing the message value. This conversion is performed using a record builder. Each record is then processed using a converter which creates the message that's published on a Kafka topic.
There are two record builders supplied with the connector, although you can write your own. The basic rule is that if you just want the message to be passed along to Kafka unchanged, the default record builder is probably the best choice. If the incoming data is in JSON format and you want to use a schema based on its structure, use the JSON record builder.
There are three converters built into Apache Kafka. You need to make sure that the incoming message format, the setting of the mq.message.body.jms configuration, the record builder and converter are all compatible. By default, everything is just treated as bytes but if you want the connector to understand the message format and apply more sophisticated processing such as single-message transforms, you'll need a more complex configuration. The following table shows the basic options that work.
| Record builder class | Incoming MQ message | mq.message.body.jms | Converter class | Outgoing Kafka message |
|---|---|---|---|---|
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | Any | false (default) | org.apache.kafka.connect.converters.ByteArrayConverter | Binary data |
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | JMS BytesMessage | true | org.apache.kafka.connect.converters.ByteArrayConverter | Binary data |
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | JMS TextMessage | true | org.apache.kafka.connect.storage.StringConverter | String data |
| com.ibm.eventstreams.connect.mqsource.builders.JsonRecordBuilder | JSON, may have schema | Not used | org.apache.kafka.connect.json.JsonConverter | JSON, no schema |
There's no single configuration that will always be right, but here are some high-level suggestions.
- Pass unchanged binary (or string) data as the Kafka message value
value.converter=org.apache.kafka.connect.converters.ByteArrayConverter- Message format is MQSTR, pass string data as the Kafka message value
mq.message.body.jms=true
value.converter=org.apache.kafka.connect.storage.StringConverter- Messages are JMS BytesMessage, pass byte array as the Kafka message value
mq.message.body.jms=true
value.converter=org.apache.kafka.connect.converters.ByteArrayConverter- Messages are JMS TextMessage, pass string data as the Kafka message value
mq.message.body.jms=true
value.converter=org.apache.kafka.connect.storage.StringConverterThe messages received from MQ are processed by a record builder which builds a Kafka Connect record to represent the message. There are two record builders supplied with the MQ source connector. The connector has a configuration option mq.message.body.jms that controls whether it interprets the MQ messages as JMS messages or regular MQ messages.
| Record builder class | mq.message.body.jms | Incoming message body | Value schema | Value class |
|---|---|---|---|---|
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | false (default) | Any | OPTIONAL_BYTES | byte[] |
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | true | JMS BytesMessage | null | byte[] |
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | true | JMS TextMessage | null | String |
| com.ibm.eventstreams.connect.mqsource.builders.DefaultRecordBuilder | true | Everything else | EXCEPTION | EXCEPTION |
| com.ibm.eventstreams.connect.mqsource.builders.JsonRecordBuilder | Not used | JSON | Depends on message | Depends on message |
You must then choose a converter than can handle the value schema and class. There are three basic converters built into Apache Kafka, with the likely useful combinations in bold.
| Converter class | Output for byte[] | Output for String | Output for compound schema |
|---|---|---|---|
| org.apache.kafka.connect.converters.ByteArrayConverter | Binary data | EXCEPTION | EXCEPTION |
| org.apache.kafka.connect.storage.StringConverter | Works, not useful | String data | Works, not useful |
| org.apache.kafka.connect.json.JsonConverter | Base-64 JSON String | JSON String | JSON data |
By default, the connector does not use keys for the Kafka messages it publishes. It can be configured to use the JMS message headers to set the key of the Kafka records. You could use this, for example, to use the MQMD correlation identifier as the partitioning key when the messages are published to Kafka. There are four valid values for the mq.record.builder.key.header that controls this behavior.
| mq.record.builder.key.header | Key schema | Key class | Recommended value for key.converter |
|---|---|---|---|
| JMSMessageID | OPTIONAL_STRING | String | org.apache.kafka.connect.storage.StringConverter |
| JMSCorrelationID | OPTIONAL_STRING | String | org.apache.kafka.connect.storage.StringConverter |
| JMSCorrelationIDAsBytes | OPTIONAL_BYTES | byte[] | org.apache.kafka.connect.converters.ByteArrayConverter |
| JMSDestination | OPTIONAL_STRING | String | org.apache.kafka.connect.storage.StringConverter |
In MQ, the message ID and correlation ID are both 24-byte arrays. As strings, the connector represents them using a sequence of 48 hexadecimal characters.
If you write your own RecordBuilder, you can access the MQMD fields of the MQ messages as JMS message properties. By default, only a subset of the MQMD fields are available, but you can get access to all of them by setting the configuration mq.message.mqmd.read. For more information, see JMS message object properties in the MQ documentation.
The connector supports authentication with user name and password and also connections secured with TLS using a server-side certificate and mutual authentication with client-side certificates. You can also choose whether to use connection security parameters (MQCSP) depending on the security settings you're using in MQ.
To enable use of TLS, set the configuration mq.ssl.cipher.suite to the name of the cipher suite which matches the CipherSpec in the SSLCIPH attribute of the MQ server-connection channel. Use the table of supported cipher suites for MQ 9.1 here as a reference. Note that the names of the CipherSpecs as used in the MQ configuration are not necessarily the same as the cipher suite names that the connector uses. The connector uses the JMS interface so it follows the Java conventions.
You will need to put the public part of the queue manager's certificate in the JSSE truststore used by the Kafka Connect worker that you're using to run the connector. If you need to specify extra arguments to the worker's JVM, you can use the EXTRA_ARGS environment variable.
You will need to put the public part of the client's certificate in the queue manager's key repository. You will also need to configure the worker's JVM with the location and password for the keystore containing the client's certificate. Alternatively, you can configure a separate keystore and truststore for the connector.
For troubleshooting, or to better understand the handshake performed by the IBM MQ Java client application in combination with your specific JSSE provider, you can enable debugging by setting javax.net.debug=ssl in the JVM environment.
The configuration options for the Kafka Connect source connector for IBM MQ are as follows:
| Name | Description | Type | Default | Valid values |
|---|---|---|---|---|
| topic | The name of the target Kafka topic | string | Topic name | |
| mq.queue.manager | The name of the MQ queue manager | string | MQ queue manager name | |
| mq.connection.mode | The connection mode - bindings or client | string | client | client, bindings |
| mq.connection.name.list | List of connection names for queue manager | string | host(port)[,host(port),...] | |
| mq.channel.name | The name of the server-connection channel | string | MQ channel name | |
| mq.queue | The name of the source MQ queue | string | MQ queue name | |
| mq.exactly.once.state.queue | The name of the MQ queue used to store state when running with exactly-once semantics | string | MQ state queue name | |
| mq.user.name | The user name for authenticating with the queue manager | string | User name | |
| mq.password | The password for authenticating with the queue manager | string | Password | |
| mq.user.authentication.mqcsp | Whether to use MQ connection security parameters (MQCSP) | boolean | true | |
| mq.ccdt.url | The URL for the CCDT file containing MQ connection details | string | URL for obtaining a CCDT file | |
| mq.record.builder | The class used to build the Kafka Connect record | string | Class implementing RecordBuilder | |
| mq.message.body.jms | Whether to interpret the message body as a JMS message type | boolean | false | |
| mq.record.builder.key.header | The JMS message header to use as the Kafka record key | string | JMSMessageID, JMSCorrelationID, JMSCorrelationIDAsBytes, JMSDestination | |
| mq.jms.properties.copy.to.kafka.headers | Whether to copy JMS message properties to Kafka headers | boolean | false | |
| mq.ssl.cipher.suite | The name of the cipher suite for TLS (SSL) connection | string | Blank or valid cipher suite | |
| mq.ssl.peer.name | The distinguished name pattern of the TLS (SSL) peer | string | Blank or DN pattern | |
| mq.ssl.keystore.location | The path to the JKS keystore to use for SSL (TLS) connections | string | JVM keystore | Local path to a JKS file |
| mq.ssl.keystore.password | The password of the JKS keystore to use for SSL (TLS) connections | string | ||
| mq.ssl.truststore.location | The path to the JKS truststore to use for SSL (TLS) connections | string | JVM truststore | Local path to a JKS file |
| mq.ssl.truststore.password | The password of the JKS truststore to use for SSL (TLS) connections | string | ||
| mq.ssl.use.ibm.cipher.mappings | Whether to set system property to control use of IBM cipher mappings | boolean | ||
| mq.batch.size | The maximum number of messages in a batch (unit of work) | integer | 250 | 1 or greater |
| mq.message.mqmd.read | Whether to enable reading of all MQMD fields | boolean | false | |
| mq.max.poll.blocked.time.ms | How long the connector will wait for the previous batch of messages to be delivered to Kafka before starting a new poll | integer | 2000 | It is important that this is less than the time defined for task.shutdown.graceful.timeout.ms as that is how long connect will wait for the task to perform lifecycle operations. |
| mq.client.reconnect.options | Options governing MQ reconnection. | string | ASDEF | ASDEF, ANY, QMGR, DISABLED |
| mq.message.receive.timeout | The timeout (in milliseconds) for receiving messages from the queue manager before returning to Kafka Connect. | long | 2000 | 1 or greater |
| mq.reconnect.delay.min.ms | The minimum delay (in milliseconds) for reconnecting to the queue manager after a connection error. | long | 64 | 1 or greater |
| mq.reconnect.delay.max.ms | The maximum delay (in milliseconds) for reconnecting to the queue manager after a connection error. | long | 8192 | 1 or greater |
Some of the connection details for MQ can be provided in a CCDT file by setting mq.ccdt.url in the MQ source connector configuration file. If using a CCDT file the mq.connection.name.list and mq.channel.name configuration options are not required.
KIP 297 introduced a mechanism to externalize secrets to be used as configuration for Kafka connectors.
Given a file mq-secrets.properties with the contents:
secret-key=passwordUpdate the worker configuration file to specify the FileConfigProvider which is included by default:
# Additional properties for the worker configuration to enable use of ConfigProviders
# multiple comma-separated provider types can be specified here
config.providers=file
config.providers.file.class=org.apache.kafka.common.config.provider.FileConfigProviderUpdate the connector configuration file to reference secret-key in the file:
mq.password=${file:mq-secret.properties:secret-key}To use a file for the mq.password in Kubernetes, you create a Secret using the file as described in the Kubernetes docs.
The MQ source connector provides at-least-once message delivery by default. This means that each MQ message will be delivered to Kafka, but in failure scenarios it is possible to have duplicated messages delivered to Kafka.
Version 2.0.0 of the MQ source connector introduced exactly-once message delivery semantics. An additional MQ queue is used to store the state of message deliveries. When exactly-once delivery is enabled all MQ messages are delivered to Kafka with no duplicated messages.
To enable exactly-once delivery, the MQ source connector must be run on Kafka Connect version 3.3.0 or later with the exactly.once.source.support property set to enabled in the Kafka connect worker configuration. See the Kafka documentation for more details about this setting, and the ACL requirements for the worker nodes.
Note: Exactly-once support for source connectors is only available in distributed mode; standalone Connect workers cannot provide exactly-once delivery semantics. Kafka Connect is in distributed mode when running the connector with Docker and when deploying the connector to Kubernetes.
To enable exactly-once delivery, the MQ source connector must be configured the mq.exactly.once.state.queue property set to the name of a pre-configured MQ queue on the same queue manager as the source MQ queue. The MQ source connector only permits a transaction.boundary property value of poll (the default value), as exactly-once delivery requires the Kafka producer transactions to be started and committed for each batch of records that the MQ source connector provides to Kafka.
Exactly-once delivery requires that only a single connector task can run in the Kafka Connect instance, hence the tasks.max property must be set to 1 to ensure that failure scenarios do not cause duplicated messages to be delivered.
Exactly-once delivery also requires that the MQ Source connector principal has a specific set of ACLs to be able to write transactionally to Kafka. See the Kafka documentation for the ACL requirements.
To start the MQ source connector with exactly-once delivery, the config/mq-source-exactly-once.json file in this repository can be used as a connector configuration template.
Note: Exactly-once delivery requires a clear state queue on start-up otherwise the connector will behave as if it is recovering from a failure state and will attempt to get undelivered messages recorded in the out-of-date state message. Therefore, ensure that the state queue is empty each time exactly-once delivery is enabled (especially if re-enabling the exactly once feature).
The following values are recommended across MQ to facilitate the exactly-once behaviour:
- On the channel used for Kafka Connect,
HBINTshould be set to 30 seconds to allow MQ transaction rollbacks to occur more quickly in failure scenarios. - On the state queue,
DEFSOPTshould be set toEXCLto ensure the state queue share option is exclusive.
Exactly-once delivery requires that messages are set to not expire and that all messages on both the source and state queue are all persistent (this is to ensure correct behaviour around queue manager restarts).
Once the MQ source connector has delivered messages to Kafka with exactly-once semantics, downstream consumers of the Kafka topic must set the isolation.level configuration property to read_committed to ensure they are only consuming transactionally committed messages.
The MQ source connector is designed to fail on start-up in certain cases to ensure that exactly-once delivery is not compromised.
In some of these failure scenarios, it will be necessary for an MQ administrator to remove messages from the exactly-once state queue before the MQ source connector can start up and begin to deliver messages from the source queue again. In these cases, the MQ source connector will have the FAILED status and the Kafka Connect logs will describe any required administrative action.
If the connector experiences a non retriable error then a ConnectException will cause the connector to go in to a FAILED state. This will require a manual restart using the Kafka Connect REST API to restart the connector.
You may receive an org.apache.kafka.common.errors.SslAuthenticationException: SSL handshake failed error when trying to run the MQ source connector using SSL to connect to your Kafka cluster. In the case that the error is caused by the following exception: Caused by: java.security.cert.CertificateException: No subject alternative DNS name matching XXXXX found., Java may be replacing the IP address of your cluster with the corresponding hostname in your /etc/hosts file. For example, to push Docker images to a custom Docker repository, you may add an entry in this file which corresponds to the IP of your repository e.g. 123.456.78.90 mycluster.icp. To fix this, you can comment out this line in your /etc/hosts file.
When configuring TLS connection to MQ, you may find that the queue manager rejects the cipher suite, in spite of the name looking correct. There are two different naming conventions for cipher suites (https://www.ibm.com/support/knowledgecenter/SSFKSJ_9.1.0/com.ibm.mq.dev.doc/q113220_.htm). Setting the configuration option mq.ssl.use.ibm.cipher.mappings=false often resolves cipher suite problems.
A commercially supported version of this connector is available for customers with a support entitlement for IBM Event Streams or IBM Cloud Pak for Integration.
For issues relating specifically to this connector, please use the GitHub issue tracker. If you do want to submit a Pull Request related to this connector, please read the contributing guide first to understand how to sign your commits.
Copyright 2017, 2020, 2023, 2024 IBM Corporation
Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.The project is licensed under the Apache 2 license.