Parallel safety of binary_upgrade_create_empty_extension

Thomas Munro <thomas.munro@enterprisedb.com>

From: Thomas Munro <thomas.munro@enterprisedb.com>
To: Pg Hackers <pgsql-hackers@postgresql.org>
Date: 2018-03-26T22:45:53Z
Lists: pgsql-hackers

Attachments

Hi,

My colleague Richard Yen came across this situation:

pg_restore: [archiver (db)] could not execute query: ERROR:  cannot
assign XIDs during a parallel operation
    Command was: -- For binary upgrade, create an empty extension and
insert objects into it
DROP EXTENSION IF EXISTS "btree_gin";
SELECT pg_catalog.binary_upgrade_create_empty_extension('btree_gin',
'public', true, '1.0', NULL, NULL, ARRAY[]::pg_catalog.text[]);

It turned out that the target cluster was running with
force_parallel_mode = on.  Here's a single character patch to mark
that function PARALLEL UNSAFE.  Obviously that'll affect only newly
initdb'd clusters after this patch, but that's what people have in a
pg_upgrade scenario.

This goes back to d89f06f0482 so I think it should probably be
back-patched to 9.6 and 10.

-- 
Thomas Munro
http://www.enterprisedb.com

Commits

  1. Fix bogus provolatile/proparallel markings on a few built-in functions.