Re: Starting new cluster from base backup

On 02/16/2015 11:31 AM, Guillaume Drolet wrote:

Dear listers,

I want to move a cluster from one machine to another. I used
pg_basebackup to create an archive and copied/extracted it over the old
PGDATA location on the new machine (the server was stopped). If I start
pgsql I get these messages in my log file:

2015-02-16 14:29:12 EST LOG:  database system was interrupted; last
known up at 2015-02-07 06:31:41 EST
2015-02-16 14:29:12 EST LOG:  invalid checkpoint record
2015-02-16 14:29:12 EST FATAL:  could not locate required checkpoint record
2015-02-16 14:29:12 EST HINT:  If you are not restoring from a backup,
try removing the file "E:/data/backup_label".
2015-02-16 14:29:12 EST LOG:  startup process (PID 3148) exited with
exit code 1
2015-02-16 14:29:12 EST LOG:  aborting startup due to startup process
failure

I assume this is due to the fact the pg_xlog folder is empty (this is
how pg_basebackup makes it in the archive) and that I haven't supplied a
recovery.conf file with the restore restore_command = 'copy
"E:\\archivedir\\%f" "%p"', and the archived WAL files.

Now my question is: it this a correct way of moving a cluster between
machines?

If yes, what WAL files will I put in pg_xlog? Would I have needed to
copy those that were in the old machine right after the base backup?

If this is not the right way to do it, what is the best way?

http://www.postgresql.org/docs/9.3/interactive/app-pgbasebackup.html

"
-X method
--xlog-method=method

    Includes the required transaction log files (WAL files) in the backup. This will include all transaction logs generated during the backup. If this option is specified, it is possible to start a postmaster directly in the extracted directory without the need to consult the log archive, thus making this a completely standalone backup ....
"

There is more under -X, so I would read the whole section.

Thanks a lot for your help,

Guillaume

--
Adrian Klaver
adrian.klaver@aklaver.com