Apache Hive v2.0.1發布

Hive是一個基于Hadoop的開源數據倉庫工具，用于存儲和處理海量結構化數據。它是非死book 2008年8月開源的一個數據倉庫框架，提供了類似于SQL語法的HQL語句作為數據訪問接口，Hive有如下優缺點：
優點：

• Hive 使用類SQL 查詢語法, 最大限度的實現了和SQL標準的兼容，大大降低了傳統數據分析人員學習的曲線；
• 使用JDBC 接口/ODBC接口，開發人員更易開發應用；
• 以MR 作為計算引擎、HDFS 作為存儲系統，為超大數據集設計的計算/ 擴展能力；
• 統一的元數據管理（Derby、MySql等），并可與Pig 、Presto 等共享；

缺點：

• Hive 的HQL 表達的能力有限，有些復雜運算用HQL 不易表達；
• 由于Hive自動生成MapReduce 作業， HQL 調優困難；
• 粒度較粗，可控性差

Hive運行架構

更新日志

Sub-task

• [HIVE-13362] - Commit binary file required for HIVE-13361

Bug修復

• [HIVE-9499] - hive.limit.query.max.table.partition makes queries fail on non-partitioned tables
• [HIVE-9862] - Vectorized execution corrupts timestamp values
• [HIVE-10729] - Query failed when select complex columns from joinned table (tez map join only)
• [HIVE-12064] - prevent transactional=false
• [HIVE-12165] - wrong result when hive.optimize.sampling.orderby=true with some aggregate functions
• [HIVE-12552] - Wrong number of reducer estimation causing job to fail
• [HIVE-12749] - Constant propagate returns string values in incorrect format
• [HIVE-12799] - Always use Schema Evolution for ACID
• [HIVE-12887] - Handle ORC schema on read with fewer columns than file schema (after Schema Evolution changes)
• [HIVE-12894] - Detect whether ORC is reading from ACID table correctly for Schema Evolution
• [HIVE-12937] - DbNotificationListener unable to clean up old notification events
• [HIVE-12990] - LLAP: ORC cache NPE without FileID support
• [HIVE-12992] - Hive on tez: Bucket map join plan is incorrect
• [HIVE-13036] - Split hive.root.logger separately to make it compatible with log4j1.x (for remaining services)
• [HIVE-13051] - Deadline class has numerous issues
• [HIVE-13056] - delegation tokens do not work with HS2 when used with http transport and kerberos
• [HIVE-13079] - LLAP: Allow reading log4j properties from default JAR resources
• [HIVE-13083] - Writing HiveDecimal to ORC can wrongly suppress present stream
• [HIVE-13086] - LLAP: Programmatically initialize log4j2 to print out the properties location
• [HIVE-13090] - Hive metastore crashes on NPE with ZooKeeperTokenStore
• [HIVE-13093] - hive metastore does not exit on start failure
• [HIVE-13105] - LLAP token hashCode and equals methods are incorrect
• [HIVE-13108] - Operators: SORT BY randomness is not safe with network partitions
• [HIVE-13110] - LLAP: Package log4j2 jars into Slider pkg
• [HIVE-13111] - Fix timestamp / interval_day_time wrong results with HIVE-9862
• [HIVE-13115] - MetaStore Direct SQL getPartitions call fail when the columns schemas for a partition are null
• [HIVE-13126] - Clean up MapJoinOperator properly to avoid object cache reuse with unintentional states
• [HIVE-13134] - JDBC: JDBC Standalone should not be in the lib dir by default
• [HIVE-13144] - HS2 can leak ZK ACL objects when curator retries to create the persistent ephemeral node
• [HIVE-13151] - Clean up UGI objects in FileSystem cache for transactions
• [HIVE-13153] - SessionID is appended to thread name twice
• [HIVE-13199] - NDC stopped working in LLAP logging
• [HIVE-13200] - Aggregation functions returning empty rows on partitioned columns
• [HIVE-13232] - Aggressively drop compression buffers in ORC OutStreams
• [HIVE-13236] - LLAP: token renewal interval needs to be set
• [HIVE-13240] - GroupByOperator: Drop the hash aggregates when closing operator
• [HIVE-13242] - DISTINCT keyword is dropped by the parser for windowing
• [HIVE-13243] - Hive drop table on encyption zone fails for external tables
• [HIVE-13255] - FloatTreeReader.nextVector is expensive
• [HIVE-13263] - Vectorization: Unable to vectorize regexp_extract/regexp_replace " Udf: GenericUDFBridge, is not supported"
• [HIVE-13285] - Orc concatenation may drop old files from moving to final path
• [HIVE-13286] - Query ID is being reused across queries
• [HIVE-13294] - AvroSerde leaks the connection in a case when reading schema from a url
• [HIVE-13296] - Add vectorized Q test with complex types showing count(*) etc work correctly
• [HIVE-13299] - Column Names trimmed of leading and trailing spaces
• [HIVE-13310] - Vectorized Projection Comparison Number Column to Scalar broken for !noNulls and selectedInUse
• [HIVE-13313] - TABLESAMPLE ROWS feature broken for Vectorization
• [HIVE-13324] - LLAP: history log for FRAGMENT_START doesn't log DagId correctly
• [HIVE-13327] - SessionID added to HS2 threadname does not trim spaces
• [HIVE-13330] - ORC vectorized string dictionary reader does not differentiate null vs empty string dictionary
• [HIVE-13346] - LLAP doesn't update metadata priority when reusing from cache; some tweaks in LRFU policy
• [HIVE-13361] - Orc concatenation should enforce the compression buffer size
• [HIVE-13379] - HIVE-12851 args do not work (slider-keytab-dir, etc.)
• [HIVE-13390] - HiveServer2: Add more test to ZK service discovery using MiniHS2
• [HIVE-13394] - Analyze table fails in tez on empty partitions/files/tables
• [HIVE-13396] - LLAP: Include hadoop-metrics2.properties file LlapServiceDriver
• [HIVE-13405] - Fix Connection Leak in OrcRawRecordMerger
• [HIVE-13428] - ZK SM in LLAP should have unique paths per cluster
• [HIVE-13463] - Fix ImportSemanticAnalyzer to allow for different src/dst filesystems
• [HIVE-13468] - branch-2 build is broken
• [HIVE-13523] - Fix connection leak in ORC RecordReader and refactor for unit testing
• [HIVE-13630] - missing license headers
• [HIVE-13645] - Beeline needs null-guard around hiveVars and hiveConfVars read

改進

• [HIVE-10115] - HS2 running on a Kerberized cluster should offer Kerberos(GSSAPI) and Delegation token(DIGEST) when alternate authentication is enabled
• [HIVE-13120] - propagate doAs when generating ORC splits
• [HIVE-13782] - Compile async query asynchronously

下載

• Source code (zip)
• Source code (tar.gz)
• 官網下載