From: | Andrey Borodin <x4mmm(at)yandex-team(dot)ru> |
---|---|
To: | Michael Paquier <michael(dot)paquier(at)gmail(dot)com> |
Cc: | Stephen Frost <sfrost(at)snowman(dot)net>, Alvaro Herrera <alvherre(at)alvh(dot)no-ip(dot)org>, pgsql-hackers <pgsql-hackers(at)postgresql(dot)org>, Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
Subject: | Re: On markers of changed data |
Date: | 2017-10-09 16:54:17 |
Message-ID: | 666B4D85-561C-47CA-8AE5-F1E628AD1468@yandex-team.ru |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Hi Michael!
> 9 окт. 2017 г., в 17:28, Michael Paquier <michael(dot)paquier(at)gmail(dot)com> написал(а):
>> VM is WAL-logged [0]
>> FSM is not [1]
>
> If you are willing to go down this road, here are my takes on the matter:
> - Any LSN map should be in a different file than FSM and VM. The VM
> uses 2 bits per blocks now, and the FSM 8 bits. The FSM is wanted to
> remain small, so trying to plug into it more data is not welcome. The
> VM bits also are dedicated to purely visibility matters. We may not
> want to use that for the VM.
> - Those should not be logged, as we would end up with tracking down
> WAL records for things that themselves track the effects of other
> records.
I was asking about FSM and VM not because I wanted to place something there, but because I was looking for the way to backup FSM and VM efficiently. VM can be copied page-incrementally, FSM cannot.
But the design you are describing resembles PTRACK[0]: fork for page changes tracking, not WAL-logged, but crash safe due to recovery from others WALs.
Best regards, Andrey Borodin.
[0] https://gist.github.com/stalkerg/ab833d94e2f64df241f1835651e06e4b <https://gist.github.com/stalkerg/ab833d94e2f64df241f1835651e06e4b>
From | Date | Subject | |
---|---|---|---|
Next Message | Tom Lane | 2017-10-09 16:58:43 | Unnecessary complexity around CurrentResourceOwner changes? |
Previous Message | Alexander Korotkov | 2017-10-09 15:15:14 | Re: Pluggable storage |