wal log recovery

SQL paper

begin;
update tab set name = 'test' where id = '1' ;
....
....
update tab set name = 'test10000000' where id = '10000000' ;

pg_sleep(300);

end;

The problem:
1 why commit soon finished, I want to make it a bit longer.?
2 if the commit time is long, the data written to disk before (if not all 10000 successful), suddenly crashes (crash), through the wal log redo,
1 millions of data in table tab can all updated?

Started by Mary at December 22, 2016 - 2:21 AM

Because the MVCC mechanism of it's sake, Postgres Crash recovery is very fast, because do not need to deal with UNDO, can this BLOG reference de brother, probably in the 0.1s

Posted by Hayden at December 31, 2016 - 3:06 AM


Thank you, see

Posted by Mary at January 06, 2017 - 3:34 AM