Horton
Menu

Re: ALTER TABLE SET ACCESS METHOD on partitioned tables

Michael Paquier <michael@paquier.xyz>

View thread
From: Michael Paquier <michael@paquier.xyz>
To: Soumyadeep Chakraborty <soumyadeep2007@gmail.com>
Cc: Zhihong Yu <zyu@yugabyte.com>, Justin Pryzby <pryzby@telsasoft.com>, pgsql-hackers <pgsql-hackers@postgresql.org>, Ashwin Agrawal <ashwinstar@gmail.com>, vanjared@vmware.com
Date: 2023-03-20T00:30:50Z
Lists: pgsql-hackers 
On Thu, Jun 09, 2022 at 11:21:58AM -0700, Soumyadeep Chakraborty wrote:
> Thanks. Fixed and rebased.

I think that I am OK with the concept of this patch to use a
partitioned table's relam as a reference when creating a partition
rather than relying on the default GUC, in a way similar to
tablespaces.

One worry I see is that forcing a recursion on the leaves on ALTER
TABLE could silently break partitions where multiple table AMs are
used across separate partitions if ALTER TABLE SET ACCESS METHOD is
used on one of the parents, though it seems like this is not something
I would much worry about as now the command is an error.

A second worry is that we would just break existing creation flows
that rely on the GUC defining the default AM.  This is worth a close
lookup at pg_dump to make sure that we do things correctly with this
patch in place..  Did you check dump and restore flows with partition
trees and --no-table-access-method?  Perhaps there should be
some regression tests with partitioned tables?

+    /*
+     * For partitioned tables, when no access method is specified, we
+     * default to the parent table's AM.
+     */
+    Assert(list_length(inheritOids) == 1);
+    /* XXX: should implement get_rel_relam? */
+    relid = linitial_oid(inheritOids);
+    tup = SearchSysCache1(RELOID, ObjectIdGetDatum(relid));
+    if (!HeapTupleIsValid(tup))
+            elog(ERROR, "cache lookup failed for relation %u", relid);

Having a wrapper for that could be useful, yes.  We don't have any
code paths that would directly need that now, from what I can see,
though.  This patch gives one reason to have one.
--
Michael

Commits

Same data as JSON: GET /api/v1/messages/:b64id/commits the thread's linked commits as JSON, with link sources. API reference →

  1. Add information about access method for partitioned relations in \dP+

  2. Fix dumps of partitioned tables with table AMs

  3. Don't use the pg_am system catalog in new test

  4. ALTER TABLE: rework determination of access method ID

  5. Allow specifying an access method for partitioned tables

  6. Review wording on tablespaces w.r.t. partitioned tables

  7. Add support for DEFAULT in ALTER TABLE .. SET ACCESS METHOD

  8. Update comment of AlterTableCmd->name in parsenodes.h