Please update your application accordingly. The default value of flag has changed from 12 to 8.- Update spark connectors with latest java clients to address CLIENT-1637 ("partition unavailable" errors occur).- Support partial updates of records using the flag.In Aerojoin API calls convert these types to long. Spark connector stores spark DateType and TimestampType as long.flag overrides Apache Spark write mode.If you are using Apache Spark 2.4.x and Aerospike Connect for Spark 2.8.0 or earlier, please plan to move to Apache Spark 3.0.x and use Aerospike Connect for Spark version 3.x.y. However, bug fixes will be available until October 14, 2022. Aerospike has ceased developing new features to support Spark 2.x.y. Therefore, 2.8.0 is the last version of Aerospike Connect for Spark 2.8 that will be compatible with that Spark branch. No more 2.x.y releases of Spark are expected, even for bug fixes. Apache Spark 2.4.8 is the last release in Spark’s 2.x.y branch.options() along with the read and write statements for the configuration to take effect. We have observed that a configuration set using "()" is not passed along to the connector, hence defaults are used by the connector, which may produce unintended results.Data Source V2 implementation for Apache Spark 3.0.0.- Extend support for Apache Spark 3.0.0 Data Source V2. Streaming update trait SupportsStreamingUpdate from Spark 3.0.0 has been renamed to SupportsStreamingUpdateAsAppend in Spark 3.1.0.Streaming write does not work with Apache Spark 3.1.0.This release does not support Aerospike 5.6 boolean bin and quota features.- Handle nulls in full record writes (REPLACE, REPLACE_ONLY, and CREATE_ONLY).- Filter out records that breach write block size in Aerospike via Spark Connector. - Spark connector with default timeout settings is timing out after 1 second.Introduced a new flag to control write throughput. - Writes are slower in the Spark Connector v2 version.Migrated from queryPartiton() call to ScanPartitions().- Support boolean bins in the Spark Connector (refer to aerospike.booleanbin in the documentation).- Data Sampling using the Spark Connector using flag.- Support batchget queries with digests in Spark Connector.- Spark connector should persist Map bins as K-Ordered.The default value of flag have changed from 12 to 8. In Aerojoin API calls convert aforementioned types to Longtype.
0 Comments
Leave a Reply. |
Details
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |