pgsql: Make DROP DATABASE command generate less WAL records.

Started by Fujii Masaoover 6 years ago4 messagescomitters
Jump to latest
#1Fujii Masao
masao.fujii@gmail.com

Make DROP DATABASE command generate less WAL records.

Previously DROP DATABASE generated as many XLOG_DBASE_DROP WAL records
as the number of tablespaces that the database to drop uses. This caused
the scans of shared_buffers as many times as the number of the tablespaces
during recovery because WAL replay of one XLOG_DBASE_DROP record needs
that full scan. This could make the recovery time longer especially
when shared_buffers is large.

This commit changes DROP DATABASE so that it generates only one
XLOG_DBASE_DROP record, and registers the information of all the tablespaces
into it. Then, WAL replay of XLOG_DBASE_DROP record needs full scan of
shared_buffers only once, and which may improve the recovery performance.

Author: Fujii Masao
Reviewed-by: Kirk Jamison, Simon Riggs
Discussion: /messages/by-id/CAHGQGwF8YwNH0ZaL+2wjZPkj+ji9UhC+Z4ScnG97WKtVY5L9iw@mail.gmail.com

Branch
------
master

Details
-------
https://git.postgresql.org/pg/commitdiff/e6d8069522c8bde8239dd1fedfb4984efa4b3a1a

Modified Files
--------------
src/backend/access/rmgrdesc/dbasedesc.c | 7 +++-
src/backend/commands/dbcommands.c | 67 +++++++++++++++++++++++----------
src/include/commands/dbcommands_xlog.h | 5 ++-
3 files changed, 56 insertions(+), 23 deletions(-)

#2Michael Paquier
michael@paquier.xyz
In reply to: Fujii Masao (#1)
Re: pgsql: Make DROP DATABASE command generate less WAL records.

On Thu, Nov 21, 2019 at 12:11:58PM +0000, Fujii Masao wrote:

Make DROP DATABASE command generate less WAL records.

Previously DROP DATABASE generated as many XLOG_DBASE_DROP WAL records
as the number of tablespaces that the database to drop uses. This caused
the scans of shared_buffers as many times as the number of the tablespaces
during recovery because WAL replay of one XLOG_DBASE_DROP record needs
that full scan. This could make the recovery time longer especially
when shared_buffers is large.

This commit changes DROP DATABASE so that it generates only one
XLOG_DBASE_DROP record, and registers the information of all the tablespaces
into it. Then, WAL replay of XLOG_DBASE_DROP record needs full scan of
shared_buffers only once, and which may improve the recovery performance.

Fujii-san, you have forgotten to bump XLOG_PAGE_MAGIC in
xlog_internal.h?
--
Michael

#3Fujii Masao
masao.fujii@gmail.com
In reply to: Michael Paquier (#2)
Re: pgsql: Make DROP DATABASE command generate less WAL records.

On Thu, Nov 21, 2019 at 9:17 PM Michael Paquier <michael@paquier.xyz> wrote:

On Thu, Nov 21, 2019 at 12:11:58PM +0000, Fujii Masao wrote:

Make DROP DATABASE command generate less WAL records.

Previously DROP DATABASE generated as many XLOG_DBASE_DROP WAL records
as the number of tablespaces that the database to drop uses. This caused
the scans of shared_buffers as many times as the number of the tablespaces
during recovery because WAL replay of one XLOG_DBASE_DROP record needs
that full scan. This could make the recovery time longer especially
when shared_buffers is large.

This commit changes DROP DATABASE so that it generates only one
XLOG_DBASE_DROP record, and registers the information of all the tablespaces
into it. Then, WAL replay of XLOG_DBASE_DROP record needs full scan of
shared_buffers only once, and which may improve the recovery performance.

Fujii-san, you have forgotten to bump XLOG_PAGE_MAGIC in
xlog_internal.h?

Thanks for pointing out that! Since the format of XLOG_DBASE_DROP WAL
record was changed, XLOG_PAGE_MAGIC must be bumped. Will do that.

Regards,

--
Fujii Masao

#4Michael Paquier
michael@paquier.xyz
In reply to: Fujii Masao (#3)
Re: pgsql: Make DROP DATABASE command generate less WAL records.

On Thu, Nov 21, 2019 at 10:03:27PM +0900, Fujii Masao wrote:

Thanks for pointing out that! Since the format of XLOG_DBASE_DROP WAL
record was changed, XLOG_PAGE_MAGIC must be bumped. Will do that.

Yep, thanks for fixing.
--
Michael