Re: Adding basic NUMA awareness

Tomas Vondra <tomas@vondra.me>

From: Tomas Vondra <tomas@vondra.me>

To: Jakub Wartak <jakub.wartak@enterprisedb.com>

Cc: Alexey Makhmutov <a.makhmutov@postgrespro.ru>, PostgreSQL Hackers <pgsql-hackers@lists.postgresql.org>, Andres Freund <andres@anarazel.de>

Date: 2025-11-21T18:49:25Z

Lists: pgsql-hackers

Attachments

v20251121-0007-NUMA-partition-PGPROC.patch (text/x-patch) patch v20251121-0007
v20251121-0006-NUMA-shared-buffers-partitioning.patch (text/x-patch) patch v20251121-0006
v20251121-0005-clock-sweep-weighted-balancing.patch (text/x-patch) patch v20251121-0005
v20251121-0004-clock-sweep-scan-all-partitions.patch (text/x-patch) patch v20251121-0004
v20251121-0003-clock-sweep-balancing-of-allocations.patch (text/x-patch) patch v20251121-0003
v20251121-0002-clock-sweep-basic-partitioning.patch (text/x-patch) patch v20251121-0002
v20251121-0001-Infrastructure-for-partitioning-shared-buf.patch (text/x-patch) patch v20251121-0001

Hi,

Here's an updated version of the patch series.

It fixes a bunch of issues in pg_buffercache_pages.c - duplicate attnums
and a incorrect array length.

The main change is in 0006 - it sets the default allocation policy for
shmem to interleaving, before doing the explicit partitioning for shared
buffers. It does it by calling numa_set_membind before the mmap(), and
then numa_interleave_memory() on the allocated shmem. It does this to
allow using MAP_POPULATE - but that's commented out by default.

This does seem to solve the SIGBUS failures for me. I still think there
might be a small chance of hitting that, because of locating an extra
"boundary" page on one of the nodes. But it should be solvable by
reserving a couple more pages.

Jakub, what do you think?


regards

-- 
Tomas Vondra

Commits

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

Sort guc_parameters.dat alphabetically by name
- fce7c73fba4e 19 (unreleased) cited