Technology Preview features are not supported with Red Hat production service-level agreements (SLAs) and might not be functionally complete; therefore, Red Hat does not recommend implementing any Technology Preview features in production environments. This Technology Preview feature provides early access to upcoming product innovations, enabling you to test functionality and provide feedback during the development process.
For more information about support scope, see link:https://access.redhat.com/support/offerings/techpreview/[Technology Preview Features Support Scope^].
{prodname}'s MongoDB connector tracks a MongoDB replica set or a MongoDB sharded cluster for document changes in databases and collections, recording those changes as events in Kafka topics.
The connector automatically handles the addition or removal of shards in a sharded cluster, changes in membership of each replica set, elections within each replica set, and awaiting the resolution of communications problems.
MongoDB's replication mechanism provides redundancy and high availability, and is the preferred way to run MongoDB in production.
MongoDB connector captures the changes in a replica set or sharded cluster.
A MongoDB _replica set_ consists of a set of servers that all have copies of the same data, and replication ensures that all changes made by clients to documents on the replica set's _primary_ are correctly applied to the other replica set's servers, called _secondaries_.
MongoDB replication works by having the primary record the changes in its _oplog_ (or operation log), and then each of the secondaries reads the primary's oplog and applies in order all of the operations to their own documents.
When a new server is added to a replica set, that server first performs an https://docs.mongodb.com/manual/core/replica-set-sync/[snapshot] of all of the databases and collections on the primary, and then reads the primary's oplog to apply all changes that might have been made since it began the snapshot.
The MongoDB connector uses this same replication mechanism, though it does not actually become a member of the replica set.
Just like MongoDB secondaries, however, the connector always reads the oplog of the replica set's primary.
And, when the connector sees a replica set for the first time, it looks at the oplog to get the last recorded transaction and then performs an _intial sync_ of the primary's databases and collections.
When all the data is copied, the connector then starts reading the oplog from the position it read earlier. Operations in the MongoDB oplog are https://docs.mongodb.com/manual/core/replica-set-oplog/[idempotent], so no matter how many times the operations are applied, they result in the same end state.
As the MongoDB connector processes the oplog, it periodically records the position in the oplog where the event originated.
When the MongoDB connector stops, it records the last oplog position that it processed, so that upon restart it simply begins reading the oplog from that position.
In other words, the connector can be stopped, upgraded or maintained, and restarted some time later, and it will pick up exactly where it left off without losing a single event.
Of course, MongoDB's oplogs are usually capped at a maximum size, which means that the connector should not be stopped for too long, or else some of the operations in the oplog might be purged before the connector has a chance to read them.
The MongoDB connector is also quite tolerant of changes in membership and leadership of the replica sets, of additions or removals of shards within a sharded cluster, and network problems that might cause communication failures.
The connector always uses the replica set's primary node to tail the oplog, so when the replica set undergoes an election and a different node becomes primary, the connector will immediately stop tailing the oplog, connect to the new primary, and start tailing the oplog using the new primary node.
Likewise, if connector experiences any problems communicating with the replica set primary, it will try to reconnect (using exponential backoff so as to not overwhelm the network or replica set) and continue tailing the oplog from where it last left off.
In this way the connector is able to dynamically adjust to changes in replica set membership and to automatically handle communication failures.
The MongoDB connector uses MongoDB's oplog to capture the changes, so the connector works only with MongoDB replica sets or with sharded clusters where each shard is a separate replica set.
See the MongoDB documentation for setting up a https://docs.mongodb.com/manual/replication/[replica set] or https://docs.mongodb.com/manual/sharding/[sharded cluster].
Also, be sure to understand how to enable https://docs.mongodb.com/manual/tutorial/deploy-replica-set-with-keyfile-access-control/#deploy-repl-set-with-auth[access control and authentication] with replica sets.
You must also have a MongoDB user that has the appropriate roles to read the `admin` database where the oplog can be read. Additionally, the user must also be able to read the `config` database in the configuration server of a sharded cluster and must have `listDatabases` privilege action.
To use the MongoDB connector with a replica set, provide the addresses of one or more replica set servers as _seed addresses_ through the connector's `mongodb.hosts` property.
The connector will use these seeds to connect to the replica set, and then once connected will get from the replica set the complete set of members and which member is primary.
The connector will start a task to connect to the primary and capture the changes from the primary's oplog.
When the replica set elects a new primary, the task will automatically switch over to the new primary.
When MongoDB is fronted by a proxy (such as with Docker on OS X or Windows), then when a client connects to the replica set and discovers the members, the MongoDB client will exclude the proxy as a valid member and will attempt and fail to connect directly to the members rather than go through the proxy.
In such a case, set the connector's optional `mongodb.members.auto.discover` configuration property to `false` to instruct the connector to forgo membership discovery and instead simply use the first seed address (specified via the `mongodb.hosts` property) as the primary node.
This may work, but still make cause issues when election occurs.
To use the MongoDB connector with a sharded cluster, configure the connector with the host addresses of the _configuration server_ replica set. When the connector connects to this replica set, it discovers that it is acting as the configuration server for a sharded cluster, discovers the information about each replica set used as a shard in the cluster, and will then start up a separate task to capture the changes from each replica set. If new shards are added to the cluster or existing shards removed, the connector will automatically adjust its tasks accordingly.
When a MongoDB connector is configured and deployed, it starts by connecting to the MongoDB servers at the seed addresses, and determines the details about each of the available replica sets.
Since each replica set has its own independent oplog, the connector will try to use a separate task for each replica set.
The connector can limit the maximum number of tasks it will use, and if not enough tasks are available the connector will assign multiple replica sets to each task, although the task will still use a separate thread for each replica set.
When running the connector against a sharded cluster, use a value of `tasks.max` that is greater than the number of replica sets.
This will allow the connector to create one task for each replica set, and will let Kafka Connect coordinate, distribute, and manage the tasks across all of the available worker processes.
The connector configuration property `mongodb.name` serves as a _logical name_ for the MongoDB replica set or sharded cluster.
The connector uses the logical name in a number of ways: as the prefix for all topic names, and as a unique identifier when recording the oplog position of each replica set.
When a task starts up using a replica set, it uses the connector's logical name and the replica set name to find an _offset_ that describes the position in the replica sets oplog where the connector previously stopped reading.
If an offset can be found and it is still in the oplog, then the task immediately proceeds with link:#tailing-the-oplog[tailing the oplog], starting at the recorded offset position.
However, if no offset is found or if the oplog no longer contains that position, the task must first obtain the current state of the replica set contents by performing a _snapshot_.
This process starts by recording the current position of the oplog and recording that as the offset (along with a flag that denotes a snapshot has been started).
The task will then proceed to copy each collection, spawning as many threads as possible (up to the value of the `initial.sync.max.threads` configuration property) to perform this work in parallel.
The connector will record a separate _read event_ for each document it sees, and that read event will contain the object's identifier, the complete state of the object, and _source_ information about the MongoDB replica set where the object was found.
Try to avoid task reassignment and reconfiguration while the connector is performing a snapshot of any replica sets. The connector does log messages with the progress of the snapshot. For utmost control, run a separate cluster of Kafka Connect for each connector.
The task will then connect to the replica set's primary node and start reading the oplog from that position, processing all of the create, insert, and delete operations and converting them into {prodname} link:#mongodb-events[change events]. Each change event includes the position in the oplog where the operation was found, and the connector periodically records this as its most recent offset. (The interval at which the offset is recorded is governed by the `offset.flush.interval.ms` {link-kafka-docs}/#connectconfigs[Kafka Connect worker configuration property].)
When the connector is stopped gracefully, the last offset processed is recorded so that, upon restart, the connector will continue exactly where it left off.
If the connector's tasks terminate unexpectedly, however, then the tasks may have processed and generated events after it last records the offset but before the last offset is recorded; upon restart, the connector begins at the last _recorded_ offset, possibly generating some the same events that were previously generated just prior to the crash.
When everything is operating nominally, Kafka consumers will actually see every message *_exactly once_*. However, when things go wrong Kafka can only guarantee consumers will see every message *_at least once_*. Therefore, your consumers need to anticipate seeing messages more than once.
====
As mentioned above, the connector tasks always use the replica set's primary node to tail the oplog, ensuring that the connector sees the most up-to-date operations as possible and can capture the changes with lower latency than if secondaries were to be used instead. When the replica set elects a new primary, the connector will immediately stop tailing the oplog, connect to the new primary, and start tailing the new primary's oplog start at the same position. Likewise, if connector experiences any problems communicating with the replica set members, it will try to reconnect (using exponential backoff so as to not overwhelm the replica set) and once connected continue tailing the oplog from where it last left off. In this way the connector is able to dynamically adjust to changes in replica set membership and to automatically handle communication failures.
The bottom line is that the MongoDB connector will continue running under most situations, though communication problems may cause the connector to wait until the problems are resolved.
The name of the Kafka topics always takes the form _logicalName_._databaseName_._collectionName_, where _logicalName_ is the link:#logical-name[logical name] of the connector as specified with the `mongodb.name` configuration property, _databaseName_ is the name of the database where the operation occurred, and _collectionName_ is the name of the MongoDB collection in which the affected document existed.
For example, consider a MongoDB replica set with an `inventory` database that contains four collections: `products`, `products_on_hand`, `customers`, and `orders`.
If the connector monitoring this database were given a logical name of `fulfillment`, then the connector would produce events on these four Kafka topics:
Be aware that Kafka only maintains total order for events written to a single topic _partition_.
Partitioning the events by key does mean that all events with the same key will always go to the same partition, ensuring that all events for a specific document are always totally ordered.
Starting with Kafka 0.10, Kafka can optionally record with the message key and value the {link-kafka-docs}.html#upgrade_10_performance_impact[_timestamp_] at which the message was created (recorded by the producer) or written to the log by Kafka.
{prodname} and Kafka Connect are designed around _continuous streams of event messages_, and the structure of these events could potentially change over time if the source of those events changed in structure or if the connector is improved or changed.
This could be difficult for consumers to deal with, so to make it very easy Kafka Connect makes each event self-contained. Every message key and value has two parts: a _schema_ and _payload_. The schema describes the structure of the payload, while the payload contains the actual data.
For a given collection, the change event's key contains a single `id` field.
Its value is the document's identifier represented as string which is derived from the https://docs.mongodb.com/manual/reference/mongodb-extended-json/[MongoDB extended JSON serialization in strict mode]. Consider a connector with a logical name of `fulfillment`, a replica set containing an `inventory` database with a `customers` collection containing documents such as:
Every change event for the `customers` collection will feature the same key structure, which in JSON looks like this:
[source,json,indent=0]
----
{
"schema": {
"type": "struct",
"name": "fulfillment.inventory.customers.Key"
"optional": false,
"fields": [
{
"field": "id",
"type": "string",
"optional": false
}
]
},
"payload": {
"id": "1004"
}
}
----
The `schema` portion of the key contains a Kafka Connect schema describing what is in the payload portion, and in our case that means that the `payload` value is not optional, is a structure defined by a schema named `fulfillment.inventory.customers.Key`, and has one required field named `id` of type `string`. If we look at the value of the key's `payload` field, we'll see that it is indeed a structure (which in JSON is just an object) with a single `id` field, whose value is a string containing the integer `1004`.
This example used a document with an integer identifier, but any valid MongoDB document identifier (including documents) will work. The value of the `id` field in the payload will simply be a string representing a MongoDB extended JSON serialization (strict mode) of the original document's `_id` field. Find below a few examples showing how `_id` fields of
different types will get encoded as the event key's payload:
The MongoDB connector ensures that all Kafka Connect _schema names_ are http://avro.apache.org/docs/current/spec.html#names[valid Avro schema names]. This means that the logical server name must start with Latin letters or an underscore (e.g., [a-z,A-Z,\_]), and the remaining characters in the logical server name and all characters in the database and collections names must be Latin letters, digits, or an underscore (e.g., [a-z,A-Z,0-9,\_]). If not, then all invalid characters will automatically be replaced with an underscore character.
This can lead to unexpected conflicts in schemas names when the logical server name, database names, and collection names contain other characters, and the only distinguishing characters between collection full names are invalid and thus replaced with underscores. The connector attempts to produce an exception in this such cases, but only when the conflicts exist between schemas used within a single connector.
* `op` is a mandatory field that contains a string value describing the type of operation. Values for the MongoDB connector are `c` for create (or insert), `u` for update, `d` for delete, and `r` for read (in the case of a snapshot).
* `after` is an optional field that if present contains the state of the document _after_ the event occurred. MongoDB's oplog entries only contain the full state of a document for _create_ events, so these are the only events that contain an _after_ field.
* `source` is a mandatory field that contains a structure describing the source metadata for the event, which in the case of MongoDB contains several fields: the {prodname} version, the logical name, the replica set's name, the namespace of the collection, the MongoDB timestamp (and ordinal of the event within the timestamp) at which the event occurred, the identifier of the MongoDB operation (e.g., the `h` field in the oplog event), and the initial sync flag if the event resulted during a snapshot.
* `ts_ms` is optional and if present contains the time (using the system clock in the JVM running the Kafka Connect task) at which the connector processed the event.
And of course, the _schema_ portion of the event message's value contains a schema that describes this envelope structure and the nested fields within it.
If we look at the `schema` portion of this event's _value_, we can see the schema for the _envelope_ is specific to the collection, and the schema for the `source` structure (which is specific to the MongoDB connector and reused across all events). Also note that the `after` value is always a string, and that by convention it will contain a JSON representation of the document.
If we look at the `payload` portion of this event's _value_, we can see the information in the event, namely that it is describing that the document was read as part of an snapshot (since `op=r` and `snapshot=true`), and that the `after` field value contains the JSON string representation of the document.
It may appear that the JSON representations of the events are much larger than the rows they describe. This is true, because the JSON representation must include the _schema_ and the _payload_ portions of the message.
ifndef::cdc-product[]
It is possible and even recommended to use the link:/docs/faq/#avro-converter[Avro Converter] to dramatically decrease the size of the actual messages written to the Kafka topics.
The value of an _update_ change event on this collection will actually have the exact same _schema_, and its payload is structured the same but will hold different values. Specifically, an update event will not have an `after` value and will instead have a `patch` string containing the JSON representation of the idempotent update operation and a `filter` string containing the JSON representation of the selection criteria for the update. The `filter` string can include multiple shard key fields for sharded collections. Here's an example:
When we compare this to the value in the _insert_ event, we see a couple of differences in the `payload` section:
* The `op` field value is now `u`, signifying that this document changed because of an update
* The `patch` field appears and has the stringified JSON representation of the actual MongoDB idempotent change to the document, which in this example involves setting the `first_name` field to a new value
Update events in MongoDB's oplog don't have the _before_ or _after_ states of the changed document, so there's no way for the connector to provide this information.
However, because _create_ or _read_ events _do_ contain the starting state, downstream consumers of the stream can actually fully-reconstruct the state by keeping the latest state for each document and applying each event to that state. {prodname} connector's are not able to keep such state, so it is not able to do this.
The value of an _delete_ event on this collection will also have the exact same _schema_, and its payload is structured the same but will hold different values.
In particular, a delete event will not have an `after` value or a `patch` value:
The MongoDB connector actually provides one other kind of event. Each _delete_ event is followed by a _tombstone_ event that has the same key but a `null` value, giving Kafka enough information to know that its https://cwiki.apache.org/confluence/display/KAFKA/Log+Compaction[Kafka log compaction] mechanism can remove _all_ messages with that key.
All MongoDB connector events are designed to work with https://cwiki.apache.org/confluence/display/KAFKA/Log+Compaction[Kafka log compaction], which allows for the removal of older messages as long as at least the most recent message for every key is kept. This is how Kafka can reclaim storage space while ensuring the topic contains a complete dataset and can be used for reloading key-based state.
If you've already installed https://zookeeper.apache.org[Zookeeper], http://kafka.apache.org/[Kafka], and {link-kafka-docs}.html#connect[Kafka Connect], then using Debezium's MongoDB connector is easy.
extract the JARs into your Kafka Connect environment, and add the directory with the JARs to Kafka Connect's plugin path using the {link-kafka-docs}/#connectconfigs[plugin.path] configuration property.
Restart your Kafka Connect process to pick up the new JARs.
Installing the MongoDB connector is a simple process whereby you only need to download the JAR, extract it to your Kafka Connect environment, and ensure the plugin's parent directory is specified in your Kafka Connect environment.
* You have link:https://zookeeper.apache.org/[Zookeeper], link:http://kafka.apache.org/[Kafka], and link:{link-kafka-docs}.html#connect[Kafka Connect] installed.
* link:https://access.redhat.com/documentation/en-us/red_hat_integration/2019-12/html-single/installing_change_data_capture_on_openshift[Installing {prodname} on OpenShift]
* link:https://access.redhat.com/documentation/en-us/red_hat_integration/2019-12/html-single/installing_change_data_capture_on_rhel[Installing {prodname} on RHEL]
If immutable containers are your thing, then check out https://hub.docker.com/r/debezium/[Debezium's Docker images] for Zookeeper, Kafka, and Kafka Connect with the MongoDB connector already pre-installed and ready to go.
Our xref:tutorial.adoc[tutorial] even walks you through using these images, and this is a great way to learn what Debezium is all about. You can even link:/blog/2016/05/31/Debezium-on-Kubernetes/[run Debezium on Kubernetes and OpenShift].
When the connector starts, it will perform a snapshot of the collections in your MongoDB replica sets and start reading the replica sets' oplogs, producing events for every inserted, updated, and deleted row.
Following is an example of the configuration for a MongoDB connector that monitors a MongoDB replica set `rs0` at port 27017 on 192.168.99.100, which we logically name `fullfillment`.
Typically, you configure the {prodname} MongoDB connector in a `.json` file using the configuration properties available for the connector.
<3> The host addresses to use to connect to the MongoDB replica set.
<4> The _logical name_ of the MongoDB replica set, which forms a namespace for generated events and is used in all the names of the Kafka topics to which the connector writes, the Kafka Connect schema names, and the namespaces of the corresponding Avro schema when the Avro Connector is used.
<5> A list of regular expressions that match the collection namespaces (for example, <dbName>.<collectionName>) of all collections to be monitored. This is optional.
endif::[]
ifdef::cdc-product[]
Following is an example of the configuration for a MongoDB connector that monitors a MongoDB replica set `rs0` at port 27017 on 192.168.99.100, which we logically name `fullfillment`.
Typically, you configure the {prodname} MongoDB connector in a `.yaml` file using the configuration properties available for the connector.
<4> The _logical name_ of the MongoDB replica set, which forms a namespace for generated events and is used in all the names of the Kafka topics to which the connector writes, the Kafka Connect schema names, and the namespaces of the corresponding Avro schema when the Avro Connector is used.
<5> A list of regular expressions that match the collection namespaces (for example, <dbName>.<collectionName>) of all collections to be monitored. This is optional.
This configuration can be sent via POST to a running Kafka Connect service, which will then record the configuration and start up the one connector task that will connect to the MongoDB replica set or sharded cluster, assign tasks for each replica set, perform a snapshot if necessary, read the oplog, and record events to Kafka topics.
The {prodname} MongoDB connector has two metric types in addition to the built-in support for JMX metrics that Zookeeper, Kafka, and Kafka Connect have.
* <<snapshot-metrics, snapshot metrics>>; for monitoring the connector when performing snapshots
|Unique name for the connector. Attempting to register again with the same name will fail. (This property is required by all Kafka Connect connectors.)
|`connector.class`
|
|The name of the Java class for the connector. Always use a value of `io.debezium.connector.mongodb.MongoDbConnector` for the MongoDB connector.
|`mongodb.hosts`
|
|The comma-separated list of hostname and port pairs (in the form 'host' or 'host:port') of the MongoDB servers in the replica set. The list can contain a single hostname and port pair. If `mongodb.members.auto.discover` is set to `false`, then the host and port pair should be prefixed with the replica set name (e.g., `rs0/localhost:27017`).
|A unique name that identifies the connector and/or MongoDB replica set or sharded cluster that this connector monitors. Each server should be monitored by at most one {prodname} connector, since this server name prefixes all persisted Kafka topics emanating from the MongoDB replica set or cluster.
|Database (authentication source) containing MongoDB credentials. This is required only when MongoDB is configured to use authentication with another authentication database than `admin`.
|Connector will use SSL to connect to MongoDB instances.
|`mongodb.ssl.invalid.hostname.allowed`
|`false`
|When SSL is enabled this setting controls whether strict hostname checking is disabled during connection phase. If `true` the connection will not prevent man-in-the-middle attacks.
|An optional comma-separated list of regular expressions that match database names to be monitored; any database name not included in the whitelist is excluded from monitoring. By default all databases is monitored.
|An optional comma-separated list of regular expressions that match database names to be excluded from monitoring; any database name not included in the blacklist is monitored. May not be used with `database.whitelist`.
|An optional comma-separated list of regular expressions that match fully-qualified namespaces for MongoDB collections to be monitored; any collection not included in the whitelist is excluded from monitoring. Each identifier is of the form _databaseName_._collectionName_. By default the connector will monitor all collections except those in the `local` and `admin` databases. May not be used with `collection.blacklist`.
|An optional comma-separated list of regular expressions that match fully-qualified namespaces for MongoDB collections to be excluded from monitoring; any collection not included in the blacklist is monitored. Each identifier is of the form _databaseName_._collectionName_. May not be used with `collection.whitelist`.
|Specifies the criteria for running a snapshot upon startup of the connector. The default is *initial*, and specifies the connector reads a snapshot when either no offset is found or if the oplog no longer contains the previous offset. The *never* option specifies that the connector should never use snapshots, instead the connector should proceed to tail the log.
|An optional comma-separated list of the fully-qualified names of fields that should be excluded from change event message values. Fully-qualified names for fields are of the form _databaseName_._collectionName_._fieldName_._nestedFieldName_, where _databaseName_ and _collectionName_ may contain the wildcard (*) which matches any characters.
|An optional comma-separated list of the fully-qualified replacements of fields that should be used to rename fields in change event message values. Fully-qualified replacements for fields are of the form _databaseName_._collectionName_._fieldName_._nestedFieldName_:__newNestedFieldName__, where _databaseName_ and _collectionName_ may contain the wildcard (*) which matches any characters, the colon character (:) is used to determine rename mapping of field. The next field replacement is applied to the result of the previous field replacement in the list, so keep this in mind when renaming multiple fields that are in the same path.
|`tasks.max`
|`1`
|The maximum number of tasks that should be created for this connector. The MongoDB connector will attempt to use a separate task for each replica set, so the default is acceptable when using the connector with a single MongoDB replica set. When using the connector with a MongoDB sharded cluster, we recommend specifying a value that is equal to or more than the number of shards in the cluster, so that the work for each replica set can be distributed by Kafka Connect.
|`initial.sync.max.threads`
|`1`
|Positive integer value that specifies the maximum number of threads used to perform an intial sync of the collections in a replica set. Defaults to 1.
| Controls whether a tombstone event should be generated after a delete event. +
When `true` the delete operations are represented by a delete event and a subsequent tombstone event. When `false` only a delete event is sent. +
Emitting the tombstone event (the default behavior) allows Kafka to completely delete all events pertaining to the given key once the source record got deleted.
|Specifies the maximum number of documents that should be read in one go from each collection while taking a snapshot.
The connector will read the collection contents in multiple batches of this size. +
Defaults to 0, which indicates that the server chooses an appropriate fetch size.
|=======================
The following _advanced_ configuration properties have good defaults that will work in most situations and therefore rarely need to be specified in the connector's configuration.
|Positive integer value that specifies the maximum size of the blocking queue into which change events read from the database log are placed before they are written to Kafka. This queue can provide backpressure to the oplog reader when, for example, writes to Kafka are slower or if Kafka is not available. Events that appear in the queue are not included in the offsets periodically recorded by this connector. Defaults to 8192, and should always be larger than the maximum batch size specified in the `max.batch.size` property.
|`max.batch.size`
|`2048`
|Positive integer value that specifies the maximum size of each batch of events that should be processed during each iteration of this connector. Defaults to 2048.
|`poll.interval.ms`
|`1000`
|Positive integer value that specifies the number of milliseconds the connector should wait during each iteration for new change events to appear. Defaults to 1000 milliseconds, or 1 second.
|`connect.backoff.initial.delay.ms`
|`1000`
|Positive integer value that specifies the initial delay when trying to reconnect to a primary after the first failed connection attempt or when no primary is available. Defaults to 1 second (1000 ms).
|`connect.backoff.max.delay.ms`
|`1000`
|Positive integer value that specifies the maximum delay when trying to reconnect to a primary after repeated failed connection attempts or when no primary is available. Defaults to 120 seconds (120,000 ms).
|`connect.max.attempts`
|`16`
|Positive integer value that specifies the maximum number of failed connection attempts to a replica set primary before an exception occurs and task is aborted. Defaults to 16, which with the defaults for `connect.backoff.initial.delay.ms` and `connect.backoff.max.delay.ms` results in just over 20 minutes of attempts before failing.
|`mongodb.members.auto.discover`
|`true`
|Boolean value that specifies whether the addresses in 'mongodb.hosts' are seeds that should be used to discover all members of the cluster or replica set (`true`), or whether the address(es) in `mongodb.hosts` should be used as is (`false`). The default is `true` and should be used in all cases except where MongoDB is link:#mongodb-replicaset[fronted by a proxy].
This will cause the oplog files to be rotated out but connector will not notice it so on restart some events are no longer available which leads to the need of re-execution of the initial snapshot.
Set this parameter to `0` to not send heartbeat messages at all. +
|`true` when connector configuration explicitly specifies the `key.converter` or `value.converter` parameters to use Avro, otherwise defaults to `false`.
{prodname} is a distributed system that captures all changes in multiple upstream databases, and will never miss or lose an event. Of course, when the system is operating nominally or being administered carefully, then {prodname} provides _exactly once_ delivery of every change event. However, if a fault does happen then the system will still not lose any events, although while it is recovering from the fault it may repeat some change events. Thus, in these abnormal situations {prodname} (like Kafka) provides _at least once_ delivery of change events.
The connector will fail upon startup, report an error/exception in the log, and stop running when the connector's configuration is invalid, or when the connector repeatedly fails to connect to MongoDB using the specified connectivity parameters. Reconnection is done using exponential backoff, and the maximum number of attempts is configurable.
In these cases, the error will have more details about the problem and possibly a suggested work around. The connector can be restarted when the configuration has been corrected or the MongoDB problem has been addressed.
Once the connector is running, if the primary node of any of the MongoDB replica sets become unavailable or unreachable, the connector will repeatedly attempt to reconnect to the primary node, using exponential backoff to prevent saturating the network or servers. If the primary remains unavailable after the configurable number of connection attempts, the connector will fail.
The attempts to reconnect are controlled by three properties:
* `connect.backoff.initial.delay.ms` - The delay before attempting to reconnect for the first time, with a default of 1 second (1000 milliseconds).
* `connect.backoff.max.delay.ms` - The maximum delay before attempting to reconnect, with a default of 120 seconds (120,000 milliseconds).
* `connect.max.attempts` - The maximum number of attempts before an error is produced, with a default of 16.
Each delay is double that of the prior delay, up to the maximum delay. Given the default values, the following table shows the delay for each failed connection attempt and the total accumulated time before failure.
If Kafka Connect is being run in distributed mode, and a Kafka Connect process is stopped gracefully, then prior to shutdown of that processes Kafka Connect will migrate all of the process' connector tasks to another Kafka Connect process in that group, and the new connector tasks will pick up exactly where the prior tasks left off.
There is a short delay in processing while the connector tasks are stopped gracefully and restarted on the new processes.
If the group contains only one process and that process is stopped gracefully, then Kafka Connect will stop the connector and record the last offset for each replica set. Upon restart, the replica set tasks will continue exactly where they left off.
=== Kafka Connect process crashes
If the Kafka Connector process stops unexpectedly, then any connector tasks it was running will terminate without recording their most recently-processed offsets.
When Kafka Connect is being run in distributed mode, it will restart those connector tasks on other processes.
However, the MongoDB connectors will resume from the last offset _recorded_ by the earlier processes, which means that the new replacement tasks may generate some of the same change events that were processed just prior to the crash.
The number of duplicate events depends on the offset flush period and the volume of data changes just before the crash.
Because there is a chance that some events may be duplicated during a recovery from failure, consumers should always anticipate some events may be duplicated. {prodname} changes are idempotent, so a sequence of events always results in the same state.
{prodname} also includes with each change event message the source-specific information about the origin of the event, including the MongoDB event's unique transaction identifier (`h`) and timestamp (`sec` and `ord`). Consumers can keep track of other of these values to know whether it has already seen a particular event.
As the connector generates change events, the Kafka Connect framework records those events in Kafka using the Kafka producer API. Kafka Connect will also periodically record the latest offset that appears in those change events, at a frequency you've specified in the Kafka Connect worker configuration. If the Kafka brokers become unavailable, the Kafka Connect worker process running the connectors will simply repeatedly attempt to reconnect to the Kafka brokers. In other words, the connector tasks will simply pause until a connection can be reestablished, at which point the connectors will resume exactly where they left off.
=== Connector is stopped for a duration
If the connector is gracefully stopped, the replica sets can continue to be used and any new changes are recorded in MongoDB's oplog.
When the connector is restarted, it will resume reading the oplog for each replica set where it last left off, recording change events for all of the changes that were made while the connector was stopped. If the connector is stopped long enough such that MongoDB purges from its oplog some operations that the connector has not read, then upon startup the connector will perform a snapshot.
A properly configured Kafka cluster is capable of https://engineering.linkedin.com/kafka/benchmarking-apache-kafka-2-million-writes-second-three-cheap-machines[massive throughput].
Kafka Connect is written with Kafka best practices, and given enough resources will also be able to handle very large numbers of database change events. Because of this, when a connector has been restarted after a while, it is very likely to catch up with the database, though how quickly will depend upon the capabilities and performance of Kafka and the volume of changes being made to the data in MongoDB.
[NOTE]
====
If the connector remains stopped for long enough, MongoDB might purge older oplog files and the connector's last position may be lost.
In this case, when the connector configured with _initial_ snapshot mode (the default) is finally restarted, the MongoDB server will no longer have the starting point and the connector will fail with an error.
====
=== MongoDB loses writes
It is possible for MongoDB to lose commits in specific failure situations. For example, if the primary applies a change and records it in its oplog before it then crashes unexpectedly, the secondary nodes may not have had a chance to read those changes from the primary's oplog before the primary crashed. If one such secondary is then elected as primary, it's oplog is missing the last changes that the old primary had recorded and no longer has those changes.
In these cases where MongoDB loses changes recorded in a primary's oplog, it is possible that the MongoDB connector may or may not capture these lost changes. At this time, there is no way to prevent this side effect of MongoDB.
