Google BigQuery

archive

true

The connector type, like 'builtin://elastic_search'

autoAck

true

false

Boolean denotes whether or not the framework will automatically acknowledge messages

className

true

The connector type’s class reference, like 'org.apache.pulsar.io.debezium.mysql.DebeziumMysqlSource'

cleanupSubscription

false

false

Boolean denotes whether the subscriptions the functions created/used should be deleted when the functions is deleted

configs

false

{}

A key/value map of config properties specific to the type of connector. See the reference table below for values.

customRuntimeOptions

false

A string that encodes options to customize the runtime, see Apache Pulsar docs for configured runtime for details

deadLetterTopic

false

Name of the dead topic where the failing messages will be sent

inputSpecs

false

The map of input topics to its consumer configuration, each configuration has schema of {"schemaType": "type-x", "serdeClassName": "name-x", "isRegexPattern": true, "receiverQueueSize": 5}

inputs

true

[]

The input topic or topics of the Sink (specified as a JSON array)

maxMessageRetries

false

Maximum number of times that a message will be redelivered before being sent to the dead letter queue

name

true

Give your sink a good name for later reference. The name must start with a lowercase alphabetic character. It can only contain lowercase alphanumeric characters, and hyphens(kebab-case).

namespace

true

The namespace you’d like the sink created under

negativeAckRedeliveryDelayMs

false

The negative ack message redelivery delay in milliseconds

parallelism

true

1

The number of sink instances to run

processingGuarantees

true

ATLEAST_ONCE

The delivery semantics applied to the Pulsar Sink. Values are 'ATLEAST_ONCE', 'ATMOST_ONCE', 'EFFECTIVELY_ONCE'

resources

false

The compute resources that need to be allocated per instance (applicable only to the process)(as a JSON string). Example: {"cpu": 0.25,"disk":1000000000,"ram":500000000}

retainKeyOrdering

false

true

Sink consumes and processes messages in key order

retainOrdering

false

false

Boolean denotes whether the Pulsar Sink consumes and processes messages in order

runtimeFlags

false

A string that encodes options to customize the runtime, see Apache Pulsar docs for configured runtime for details

secrets

false

This is a map of secretName(that is how the secret is going to be accessed in the function via context) to an object that encapsulates how the secret is fetched by the underlying secrets provider. The type of an value here can be found by the SecretProviderConfigurator.getSecretObjectType() method

sourceSubscriptionName

false

Pulsar source subscription name if user wants a specific subscription-name for input-topic consumer

sourceSubscriptionPosition

false

Earliest

The position to begin reading from the source

tenant

true

The tenant you’d like the sink created under

timeoutMs

false

5000

Denotes the message timeout in milliseconds

topicToSchemaProperties

false

topicToSchemaType

false

The map of input topics to Schema types or class names (specified as a JSON object)

topicToSerdeClassName

false

The map of input topics to SerDe class names (specified as a JSON object)

topicsPattern

false

TopicsPattern to consume from list of topics under a namespace that match the pattern. [input] and [topicsPattern] are mutually exclusive. Add SerDe class name for a pattern in customSerdeInputs (supported for java fun only)

kafkaConnectorSinkClass

true

A Kafka-connector sink class to use. Unless you’ve developed your own, use the value "com.wepay.kafka.connect.bigquery.BigQuerySinkConnector".

offsetStorageTopic

true

Pulsar topic to store offsets at. This is an additional topic to your topic with the actual data going to BigQuery.

sanitizeTopicName

true

Some connectors cannot handle Pulsar topic names like persistent://a/b/topic and do not sanitize the topic name themselves. If enabled, all non alpha-digital characters in topic name will be replaced with underscores. In some cases this may result in topic name collisions (topic_a and topic.a will become the same)

This value MUST be set to true. Any other value will result in an error.

batchSize

false

16384

Size of messages in bytes the sink will attempt to batch messages together before flush.

collapsePartitionedTopics

false

false

Supply Kafka record with topic name without -partition- suffix for partitioned topics.

kafkaConnectorConfigProperties

false

{}

A key/value map of config properties to pass to the Kafka connector. See the reference table below.

lingerTimeMs

false

2147483647L

Time interval in milliseconds the sink will attempt to batch messages together before flush.

maxBatchBitsForOffset

false

12

Number of bits (0 to 20) to use for index of message in the batch for translation into an offset. 0 to disable this behavior (Messages from the same batch will have the same offset which can affect some connectors.)

topic

true

The Kafka topic name that is passed to the Kafka sink.

unwrapKeyValueIfAvailable

false

true

In case of Record<KeyValue<>> data use key from KeyValue<> instead of one from Record.

useIndexAsOffset

false

true

Allows use of message index instead of message sequenceId as offset, if available. Requires AppendIndexMetadataInterceptor and exposingBrokerEntryMetadataToClientEnabled=true on brokers.

useOptionalPrimitives

false

false

Pulsar schema does not contain information whether the Schema is optional, Kafka’s does. This provides a way to force all primitive schemas to be optional for Kafka.

allBQFieldsNullable

false

false

If true, no fields in any produced BigQuery schema are REQUIRED. All non-nullable Avro fields are translated as NULLABLE (or REPEATED, if arrays).

allowBigQueryRequiredFieldRelaxation

false

false

If true, fields in BigQuery Schema can be changed from REQUIRED to NULLABLE.

allowNewBigQueryFields

false

false

If true, new fields can be added to BigQuery tables during subsequent schema updates.

allowSchemaUnionization

false

false

If true, the existing table schema (if one is present) will be unionized with new record schemas during schema updates. If false, the record of the last schema in a batch will be used for any necessary table creation and schema update attempts.

Setting allowSchemaUnionization to false and allowNewBigQueryFields and allowBigQueryRequiredFieldRelaxation to true is equivalent to setting autoUpdateSchemas to true in older (pre-2.0.0) versions of this connector.

In this case, if BigQuery raises a schema validation exception or a table doesn’t exist when a writing a batch, the connector will try to remediate by required field relaxation and/or adding new fields.

If allowSchemaUnionization, allowNewBigQueryFields, and allowBigQueryRequiredFieldRelaxation are true, the connector will create or update tables with a schema whose fields are a union of the existing table schema’s fields and the ones present in all of the records of the current batch.

The key difference is that with unionization disabled, new record schemas have to be a superset of the table schema in BigQuery.

allowSchemaUnionization is a useful tool to make things work. For example, if you’d like to remove fields from data upstream, the updated schemas still work in the connector. It is similarly useful when different tasks see records whose schemas contain different fields that are not in the table. However, note with caution that if allowSchemaUnionization is set and some bad records are in the topic, the BigQuery schema may be permanently changed. This presents two issues: first, since BigQuery doesn’t allow columns to be dropped from tables, they’ll add unnecessary noise to the schema. Second, since BigQuery doesn’t allow column types to be modified, they could completely break pipelines down the road where well-behaved records have schemas whose field names overlap with the accidentally-added columns in the table, but use a different type.

autoCreateBucket

false

true

Whether to automatically create the given bucket, if it does not exist.

autoCreateTables

false

false

Automatically create BigQuery tables if they don’t already exist

avroDataCacheSize

false

100

The size of the cache to use when converting schemas from Avro to Kafka Connect.

batchLoadIntervalSec

false

120

The interval, in seconds, in which to attempt to run GCS to BigQuery load jobs. Only relevant if enableBatchLoad is configured.

bigQueryMessageTimePartitioning

false

false

Whether or not to use the message time when inserting records. Default uses the connector processing time.

bigQueryPartitionDecorator

false

true

Whether or not to append partition decorator to BigQuery table name when inserting records. Default is true. Setting this to true appends partition decorator to table name (e.g. table$yyyyMMdd depending on the configuration set for bigQueryPartitionDecorator). Setting this to false bypasses the logic to append the partition decorator and uses raw table name for inserts.

bigQueryRetry

false

0

The number of retry attempts made for a BigQuery request that fails with a backend error or a quota exceeded error.

bigQueryRetryWait

false

1000

The minimum amount of time, in milliseconds, to wait between retry attempts for a BigQuery backend or quota exceeded error.

clusteringPartitionFieldNames

false

Comma-separated list of fields where data is clustered in BigQuery.

convertDoubleSpecialValues

false

false

Designates whether +Infinity is converted to Double.MAX_VALUE and whether -Infinity and NaN are converted to Double.MIN_VALUE to ensure successfull delivery to BigQuery.

defaultDataset

true

The default dataset to be used

deleteEnabled

false

false

Enable delete functionality on the connector through the use of record keys, intermediate tables, and periodic merge flushes. A delete will be performed when a record with a null value (that is–a tombstone record) is read. This feature will not work with SMTs that change the name of the topic.

enableBatchLoad

false

“”

Beta Feature Use with caution. The sublist of topics to be batch loaded through GCS.

gcsBucketName

false

"”

The name of the bucket where Google Cloud Storage (GCS) blobs are located. These blobs are used to batch-load to BigQuery. This is applicable only if enableBatchLoad is configured.

includeKafkaData

false

false

Whether to include an extra block containing the Kafka source topic, offset, and partition information in the resulting BigQuery rows.

intermediateTableSuffix

false

“.tmp”

A suffix that will be appended to the names of destination tables to create the names for the corresponding intermediate tables. Multiple intermediate tables may be created for a single destination table, but their names will always start with the name of the destination table, followed by this suffix, and possibly followed by an additional suffix.

kafkaDataFieldName

false

The Kafka data field name. The default value is null, which means the Kafka Data field will not be included.

kafkaKeyFieldName

false

The Kafka key field name. The default value is null, which means the Kafka Key field will not be included.

keyfile

true

Can be either a string representation of the Google credentials file or the path to the Google credentials file itself.

When using the Astra Streaming UI, the string representation must be used. If using pulsar-admin with Astra Streaming, either the representation or file can be used.

keySource

true

FILE

Determines whether the keyfile configuration is the path to the credentials JSON file or to the JSON itself. Available values are FILE and JSON.

When using the Astra Streaming UI, JSON will be the only option. If using pulsar-admin with Astra Streaming, either the representation or file can be used.

name

true

The name of the connector. Use the same value as Pulsar sink name.

mergeIntervalMs

false

60_000L

How often (in milliseconds) to perform a merge flush, if upsert/delete is enabled. Can be set to -1 to disable periodic flushing.

mergeRecordsThreshold

false

-1

How many records to write to an intermediate table before performing a merge flush, if upsert/delete is enabled. Can be set to -1 to disable record count-based flushing.

project

true

The BigQuery project to write to

queueSize

false

-1

The maximum size (or -1 for no maximum size) of the worker queue for BigQuery write requests before all topics are paused. This is a soft limit; the size of the queue can go over this before topics are paused. All topics resume once a flush is triggered or the size of the queue drops under half of the maximum size.

sanitizeTopics

true

false

Designates whether to automatically sanitize topic names before using them as table names. If not enabled, topic names are used as table names.

The only accepted value is false. Providing any other value will result in an error.

schemaRetriever

false

com.wepay.kafka.connect.bigquery.retrieve.IdentitySchemaRetriever

A class that can be used for automatically creating tables and/or updating schemas.

threadPoolSize

false

10

The size of the BigQuery write thread pool. This establishes the maximum number of concurrent writes to BigQuery.

timePartitioningType

false

DAY

The time partitioning type to use when creating tables. Existing tables will not be altered to use this partitioning type. Valid Values: (case insensitive) [MONTH, YEAR, HOUR, DAY]

timestampPartitionFieldName

false

The name of the field in the value that contains the timestamp to partition by in BigQuery and enable timestamp partitioning for each table. Leave this configuration blank, to enable ingestion time partitioning for each table.

topic2TableMap

false

Map of topics to tables (optional).

Format: comma-separated tuples, e.g. <topic-1>:<table-1>,<topic-2>:<table-2>,..

Note, because sanitizeTopicName must be true, that in-turn means any alphanumeric character in the topic name will be replaced as underscore “_”. So when creating a mapping you need to take the underscores into account.

For example, if the topic name is provided as “persistent://a/b/c-d” then the mapping topic name would be “persistent___a_b_c_d”.

topics,true,,"A list of Kafka topics to read from. Use the same name as the Pulsar topic (not the whole address, just the topic name).

upsertEnabled

false

false

Enable upsert functionality on the connector through the use of record keys, intermediate tables, and periodic merge flushes. Row-matching will be performed based on the contents of record keys. This feature won’t work with SMTs that change the name of the topic.