I'm testing PITR using pgbench and postgresql ver.8.0bata.
I think that result of recovery is wrong. My test procedure
is as follows:
--------------------
I edited postgresql.conf for PITR and started the postmaster.
And I executed "pgbench -t 20000". % pgbench -t 20000
I did backup procedure before end of pgbench. % psql -c "SELECT pg_start_backup('label1')" % cd $PGDATA % tar cf
/tmp/back.tar./* % psql -c "SELECT pg_stop_backup()"
restore and recovery % pg_ctl stop % cp -R $PGDATA/pg_xlog /tmp/. % rm -rf $PGDATA/* % cd $PGDATA % tar xf
/tmp/data.tar
create "recovery.conf" file
% rm -rf $PGDATA/pg_xlog % cp -R /tmp/pg_xlog $PGDATA/. % pg_ctl start
check data after recovery % psql -c "SELECT count(*) from history" count ------- 19999 (1 row)
------------------------
Number of records should be 20000, but result is 19999.
I found lack of data that was inserted near backup time, as a
result of comparing original "history" table and "history" table which
was recovered.
Is my backup procedure wrong?
Thanks,
Satoshi OKADA