Re[2]: [HACKERS] standby, pg_basebackup and last xlog file

Hello all and Heikki personally

Thank you for your answer

I have some new points:

Понедельник, 21 января 2013, 10:08 +02:00 от Heikki Linnakangas <hlinnakangas(at)vmware(dot)com>:
>On 21.01.2013 09:14, Миша Тюрин wrote:
>> Is there any reason why pg_basebackup has limitation in an online backup from the standby: "The backup history file is not created in the database cluster backed up." ?
>
>WAL archiving isn't active in a standby, so even if it created a backup
>history file, it wouldn't go anywhere. Also, the way the backup history
>files are named, if you take a backup on the master at the same time (or
>another backup at the same time in the standby), you would end up with
>multiple backup history files with the same name.
>
>> So i can't get last xlog file needed to restore :(
>
>Good point. That was an oversight in the patch that allowed base backups
>from a standby.
>
>> Also maybe i can use something like ( pg_last_xlog_replay_location() + 1 ) after pg_basebackup finished.
>
>Yeah, that should work.
>
>> Does anybody know true way to getting last xlog file in case of applying pg_basebackup to standby?
>> How pg_basebackup gets last xlog file in case of standby and -x option?
>
>The server returns the begin and end WAL locations to pg_basebackup,
>pg_basebackup just doesn't normally print them out to the user. In
>verbose mode, it does actually print them out, but only with -x, so that
>doesn't help you either. If you can compile from source, you could
>modify pg_basebackup.c to print the locations without -x and --verbose,
>search lines that print out "transaction log start point / end position".

1) we can get last xlog by using control data's "Minimum recovery ending location"

>
>
>As a workaround, without modifying the source, you can do this after
>pg_basebackup has finished, to get the first WAL file you need:
>
>$ pg_controldata backupdir | grep "REDO WAL file"
>Latest checkpoint's REDO WAL file: 000000030000000000000009

and I would like to correct your suggestion about first wal file
2.1)  we should use backup_label to determine first needed wal
2.2)  and we must not use redo from checkpoint. because there are might be more than one checkpoint during base_backup

>
>And as you suggested, pg_last_xlog_replay_location() for the last WAL
>file you need.
>
>- Heikki
>
>
>--
>Sent via pgsql-hackers mailing list (pgsql-hackers(at)postgresql(dot)org)
>To make changes to your subscription:
>http://www.postgresql.org/mailpref/pgsql-hackers

- Misha