Hello!
I have two PSQL nodes and I have problems trying to get a data replica from 1 node to another.
There are no PSQL connections to any node.
I do PITR from node 1 to 2:
- Stop PSQL from node 2.
- Reload node 1 with postgresql.conf:
- wal_level = archive
- archive_mode = on
- archive_command = 'test ! -f /var/lib/postgresql/pg_xlog_archive/%f && /bin/cp %p /var/lib/postgresql/pg_xlog_archive/%f''
- Run pg_start_backup('label',true)
- Create 'recovery.conf' with 'restore_command = 'scp psql-node1:/var/lib/postgresql/pg_xlog_archive/%f %p'
- tar.gz all PSQL 1 folders
- Run pg_stop_backup()
- Reload node 1 with archive_comand = 'exit 0'
- Move tar.gz from node 1 to node 2
- Delete 'recovery.conf' from node 1
- Run pg_switch_xlog() on node 1
- Untar tar.gz in node 2
- Start PSQL in node 2
- Node 2 starts recovery phase...
- Restore process finishes.
But when I create a new database or tables, they get different OID in both nodes.
I checked PSQL logs during backup/restore and I don't see antyhing weird, except that node 2 starts with a different TimelineID (+1 in relation to node 1).
However, if run the backup/restore process stopping PSQL 1 at first, then start it, and then run backup/restore process, then I get same data in both nodes and "nextOID" are also the same.
It seems like there is data somewhere that is not copied to the files unless I restart the PSQL service.
Is it the expected behaviour?
I tried with different environments:
- PSQL 9.1 (different versions) and with 9.3.5
- Debian packages and Suse.
