Horton

Re: [Proposal] Fully WAL logged CREATE DATABASE - No Checkpoints

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Justin Pryzby <pryzby@telsasoft.com>
Cc: Dilip Kumar <dilipbalaut@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, Ashutosh Sharma <ashu.coek88@gmail.com>, Maciek Sakrejda <m.sakrejda@gmail.com>, Bruce Momjian <bruce@momjian.us>, Alvaro Herrera <alvherre@alvh.no-ip.org>, Andrew Dunstan <andrew@dunslane.net>, Heikki Linnakangas <hlinnaka@iki.fi>, pgsql-hackers@lists.postgresql.org
Date: 2022-08-03T18:26:43Z
Lists: pgsql-hackers 
Hi,

On 2022-08-03 12:01:18 -0500, Justin Pryzby wrote:
> Now, I've reproduced the problem under valgrind, but it doesn't show anything
> useful

Yea, that looks like an issue on a different level.

> 
> pryzbyj@pryzbyj:~$ while :; do psql -h /tmp template1 -c "DROP DATABASE a" -c "CREATE DATABASE a TEMPLATE postgres STRATEGY wal_log"; done
> ERROR:  database "a" does not exist
> CREATE DATABASE
> ^CCancel request sent
> ERROR:  canceling statement due to user request
> ERROR:  database "a" already exists
> ^C

Hm. This looks more like an issue of DROP DATABASE not being interruptible. I
suspect this isn't actually related to STRATEGY wal_log and could likely be
reproduced in older versions too.

It's pretty obvious that dropdb() isn't safe against being interrupted. We
delete the data before we have committed the deletion of the pg_database
entry.

Seems like we should hold interrupts across the remove_dbtablespaces() until
*after* we've committed the transaction?

Greetings,

Andres Freund

Commits

  1. When using the WAL-logged CREATE DATABASE strategy, bulk extend.

  2. Avoid using a fake relcache entry to own an SmgrRelation.

  3. Fix data-corruption hazard in WAL-logged CREATE DATABASE.

  4. initdb: When running CREATE DATABASE, use STRATEGY = WAL_COPY.

  5. Simplify a needlessly-complicated regular expression.

  6. In 020_createdb.pl, change order of command-line arguments.

  7. Add new block-by-block strategy for CREATE DATABASE.

  8. Fix replay of create database records on standby

  9. Refactor code for reading and writing relation map files.

  10. Replace RelationOpenSmgr() with RelationGetSmgr().

  11. Refactor the fsync queue for wider use.