事务实现原理和Write Ahead Logging（WAL） ● 分布式事务和两阶段提交的原理 ● Greenplum两阶段提交协议的实现 ● Greenplum两阶段提交协议的优化 Outline 7 事务的属性：ACID 属性 含义 数据库系统的实现 Atomic 原子性 事务中的操作要么全部正确执行，要么完全不 执行。 Write Ahead Logging，分布式事务：两阶段提交协议 两阶段加锁（Two Phase Locking, 2PL）、乐观并发控制 （OCC） Durability 持久性 一个事务在提交之后，该事务对数据库的改变 是持久的。 Write Ahead Logging + 存储管理 Jim Gray于1981年VLDB描述了事务的原子性、一致性和持久性，在此基础上，Haerder和Reuter在1983年中提出了事务的隔离性并提出术语 “ACID”，自此，事务的ACID四个性质成为业内标准术语 Pool未提交事务所修改的脏页刷回到持久存储，为了保证原子性， 先把Undo Log写入日志文件。Undo Log记录修改数据对象的旧值(Before Image, BFIM) Solution: Logging 15 缓冲区管理策略和事务恢复的关系 Force No-Force Steal Undo / No-Redo Undo + Redo (performance: fastest

the Greenplum Virtual Appliance 378 Importing and Deploying the Greenplum Virtual Appliance 378 Logging In To The Appliance Virtual Machine 379 Deleting or Removing a Greenplum Virtual Appliance 379 instances 637 Example 3—Multiple gpfdists instances 637 Example 4—Single gpfdist instance with error logging 637 Example 5—TEXT Format on a Hadoop Distributed File Server 638 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 979 Using JDBC 980 Exception Handling 980 Savepoints 981 Logging 981 Security 981 Installation 982 Trusted Language 982 Some PL/Java Issues and Solutions 982

the Greenplum Virtual Appliance 361 Importing and Deploying the Greenplum Virtual Appliance 361 Logging In To The Appliance Virtual Machine 362 Deleting or Removing a Greenplum Virtual Appliance 362 instances 608 Example 3—Multiple gpfdists instances 608 Example 4—Single gpfdist instance with error logging 608 Example 5—TEXT Format on a Hadoop Distributed File Server 609 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 947 Using JDBC 948 Exception Handling 949 Savepoints 949 Logging 950 Security 950 Installation 950 Trusted Language 950 Some PL/Java Issues and Solutions 951

instances 630 Example 3—Multiple gpfdists instances 631 Example 4—Single gpfdist instance with error logging 631 Example 5—TEXT Format on a Hadoop Distributed File Server 632 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 966 Using JDBC 967 Exception Handling 967 Savepoints 968 Logging 968 Security 969 Installation 969 Trusted Language 969 Some PL/Java Issues and Solutions 969 1127 Avoiding an Extreme Skew Warning 1128 Checking for and Terminating Overflowed Backends 1128 Logging Statements that Cause Overflowed Subtransactions 1129 Viewing Metadata Information about Database

instances 608 Example 3—Multiple gpfdists instances 608 Example 4—Single gpfdist instance with error logging 608 Example 5—TEXT Format on a Hadoop Distributed File Server 609 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 936 Using JDBC 937 Exception Handling 937 Savepoints 938 Logging 938 Security 938 Installation 938 Trusted Language 939 Some PL/Java Issues and Solutions 939 1093 Avoiding an Extreme Skew Warning 1094 Checking for and Terminating Overflowed Backends 1094 Logging Statements that Cause Overflowed Subtransactions 1095 Viewing Metadata Information about Database

instances 605 Example 3—Multiple gpfdists instances 605 Example 4—Single gpfdist instance with error logging 606 Example 5—TEXT Format on a Hadoop Distributed File Server 606 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 931 Using JDBC 932 Exception Handling 932 Savepoints 933 Logging 933 Security 934 Installation 934 Trusted Language 934 Some PL/Java Issues and Solutions 934 1089 Avoiding an Extreme Skew Warning 1089 Checking for and Terminating Overflowed Backends 1090 Logging Statements that Cause Overflowed Subtransactions 1091 Viewing Metadata Information about Database

instances 444 Example 3—Multiple gpfdists instances 444 Example 4—Single gpfdist instance with error logging 444 VMware Greenplum 7 Documentation VMware by Broadcom 18 Example 5—TEXT Format on a Hadoop Distributed Docker References 746 Using PL/Container 746 PL/Container Resource Management 746 PL/Container Logging 747 PL/Container Function Limitations 747 Developing PL/Container functions 748 Basic Function Using the ResultSetHandle Interface 768 Using JDBC 769 Exception Handling 769 Savepoints 770 Logging 770 Security 770 Installation 771 Trusted Language 771 Some PL/Java Issues and Solutions 771

instances 518 Example 3—Multiple gpfdists instances 518 Example 4—Single gpfdist instance with error logging 519 Example 5—TEXT Format on a Hadoop Distributed File Server 519 Example 6—Multiple files in CSV Using the ResultSetHandle Interface 814 Using JDBC 815 Exception Handling 815 Savepoints 815 Logging 815 Security 816 Installation 816 VMware Tanzu Greenplum 6 Documentation VMware, Inc 39 Trusted Other Postgres Planner Configuration Parameters 1569 Query Plan Execution 1569 Error Reporting and Logging Parameters 1569 Log Rotation 1569 When to Log 1569 What to Log 1570 System Monitoring Parameters

ies|Log_rotation|Log_line_prefix|logging_collector|log_filen ame|log_duration" # requires logging_collector to be on. logging_collector = on # # Enable capturing of stderr and csvlog # These are only used if logging_collector is on: log_directory = 'pg_log' # directory where log files are written, log_filename = 'postgresql-%a # supported by the operating system: # ERROR REPORTING AND LOGGING # %r = remote host and port 2018年PostgreSQL中国技术大会 微信号：laohouzi999

Exception Handling 451 Savepoints 452 VMware Greenplum v6.19 Documentation VMware, Inc. 22 Logging 452 Security 0 Installation 452 Trusted Language 453 Some PL/Java Issues and Solutions 453 Multi-threading gpfdists instances 0 Example 4—Single gpfdist instance with error logging 807 Example 4—Single gpfdist instance with error logging 0 Example 5—TEXT Format on a Hadoop Distributed File Server 807 Example Other Postgres Planner Configuration Parameters 1558 Query Plan Execution 1558 Error Reporting and Logging Parameters 1558 Log Rotation 1558 When to Log 1558 What to Log 1559 System Monitoring Parameters