RE: Re: Logging the feature of SQL-level read/write commits

2019-05-05 Thread legrand legrand
Hi, good point ! wal2Json seems to correspond to your needs, this is first designed for Change Data Capture, taht could generate a (very) big size of logs. You didn't tell us much about your use case ... and maybe, if the number of data modifications is not too big, and the number of tables to

RE: Re: Logging the feature of SQL-level read/write commits

2019-05-05 Thread Ronny Ko
Hi Legrand & Bruce, Thanks for your  thoughts. I think MariaDB's temporal queries could work.  If I want to use it in PostgreSQL, I could use Logical Decoding plugins for recording all INSERT/UPDATE/DELETE history of each table: - https://debezium.io/docs/install/postgres-plugins/ - wal2json

Re: Logging the feature of SQL-level read/write commits

2019-05-05 Thread Bruce Momjian
On Sat, May 4, 2019 at 02:32:06PM +0900, Ronny Ko wrote: > Hi, > > I am trying to efficiently rollback a manually selectedd subset of committed > SQL transactions by scanning an SQL transaction log. This feature is useful > when a database administrator wants to rollback not the entire database

Re: Logging the feature of SQL-level read/write commits

2019-05-05 Thread legrand legrand
Hello, may be you can find more informations regarding WAL concepts in Write Ahead Logging — WAL http://www.interdb.jp/pg/pgsql09.html It seems very complicated to change WAL format ... Maybe there are other solutions to answer your need, I found many interesting solutions in postgres archives

Re: Logging the feature of SQL-level read/write commits

2019-05-04 Thread Tomas Vondra
On Sat, May 04, 2019 at 02:32:06PM +0900, Ronny Ko wrote: Hi, I am trying to efficiently rollback a manually selectedd subset of comm

Logging the feature of SQL-level read/write commits

2019-05-03 Thread Ronny Ko
Hi,I am trying to efficiently rollback a manually selectedd subset of committed SQL transactions by scanning an SQL transaction log. This feature is useful when a database administrator wants to rollback not the entire database system, but only particular SQL statements that affect a certai