13.1. Release 337 (25 Jun 2020)#
Note
This release fixes a potential security vulnerability when secure internal communication is enabled in a cluster. A malicious attacker can take advantage of this vulnerability to escalate privileges to internal APIs. We encourage everyone to upgrade as soon as possible.
General Changes#
- Fix incorrect results for inequality join involving
NaN. (#4120) - Fix peak non-revocable memory metric in event listener. (#4096)
- Fix queued query JMX stats. (#4129)
- Fix rendering of types in the output of
DESCRIBE INPUT. (#4023) - Improve performance of queries involving comparisons between
DOUBLEorREALvalues and integer values. (#3533) - Reduce idle CPU consumption in coordinator. (#3990)
- Add peak non-revocable memory metric to query stats. (#4096)
- Add support for variable-precision
TIMESTAMP WITH TIME ZONEtype (#3947) - Add support for
INpredicate with subqueries in outer join condition. (#4151) - Add support for quantified comparisons (e.g.,
> ALL (...)) in aggregation queries. (#4128) - Add Druid Connector. (#3522)
- Add
translate()function. (#4080) - Reduce worker graceful shutdown duration. (#4192)
Security Changes#
- Disable insecure authentication over HTTP by default when HTTPS with authentication is enabled. This
can be overridden via the
http-server.authentication.allow-insecure-over-httpconfiguration property. (#4199) - Add support for insecure authentication over HTTPS to the Web UI. (#4199)
- Add System Information Rules which control the ability of a user to access to read and write system management information. (#4199)
- Disable user impersonation in default system security. (#4082)
Hive Connector Changes#
- Fix reading CSV tables with
separatorChar,quoteCharorescapeChartable property containing more than one character. For compatibility with Hive, only first character is considered and remaining are ignored. (#3891) - Improve performance of
INSERTqueries writing to bucketed tables when some buckets do not contain any data. (#1375) - Improve performance of queries reading Parquet data with predicates on
timestampcolumns. (#4104) - Improve performance for join queries over partitioned tables. (#4156)
- Add support for
null_formattable property for tables using TextFile storage format (#4056) - Add support for
null_formattable property for tables using RCText and SequenceFile storage formats (#4143) - Add optimized Parquet writer. The new writer is disabled by default, and can be enabled with the
parquet_optimized_writer_enabledsession property or thehive.parquet.optimized-writer.enabledconfiguration property. (#3400) - Add support caching data in Azure Data Lake and AliyunOSS storage. (#4213)
- Fix failures when caching data from Google Cloud Storage. (#4213)
- Support ACID data files naming used when direct inserts are enabled in Hive (HIVE-21164). Direct inserts is an upcoming feature in Hive 4. (#4049)