Re: [HACKERS] On markers of changed data

Hello!


> 9 окт. 2017 г., в 10:23, Andrey Borodin <x4mmm@yandex-team.ru> написал(а):
>
> Thanks, Stephen, this actually pointed what to look for
> VM is WAL-logged [0]
> FSM is not [1]
>
> [0]
https://github.com/postgres/postgres/blob/113b0045e20d40f726a0a30e33214455e4f1385e/src/backend/access/heap/visibilitymap.c#L315
> [1]
https://github.com/postgres/postgres/blob/1d25779284fe1ba08ecd57e647292a9deb241376/src/backend/storage/freespace/freespace.c#L593

After tests of binary equivalence before and after backup I've come to conclusion, that Visibility Map cannot be
backupedincrementally: it's bits are unset without page LSN bump. This can lead to wrong results of Index Only Scans
executedon freshly restored backups. 

In my implementation of incremental backup in WAL-G I will disable any FSM, VM and XACT\CLOG incrementation.

Posting this for the record, so that if someone goes this way info will be available. Thank you for your attention.

Best regards, Andrey Borodin.

--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers