Re: Hash index inserts serialise on the metapage tuple counter

From: Greg Burd <greg(at)burd(dot)me>
To: geidav(dot)pg(at)gmail(dot)com
Cc: pgsql-hackers(at)lists(dot)postgresql(dot)org, michael(at)paquier(dot)xyz
Subject: Re: Hash index inserts serialise on the metapage tuple counter
Date: 2026-11-20 14:21:53
Message-ID: 179517850046.3245040.1043782317300930601@burd.me
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

v5 attached, rebased onto 4a26e36f87f.

0002: _hash_freeovflpage() stopped writing the metapage in v3, but the
redo code for XLOG_HASH_SQUEEZE_PAGE still had a branch that replays it
as block 6. That branch is removed, along with the block-6 line in the
record's comment, and HASH_XLOG_FREE_OVFL_BUFS drops from 6 to 5. The
record format was already changing for master only, so there is nothing
left to keep compatible.

0001 is unchanged.

-greg

Attachment Content-Type Size
v5-0002-Allocate-and-free-hash-overflow-pages-without-the.patch text/x-patch 43.2 KB
v5-0001-Don-t-make-hash-index-inserts-serialize-on-the-me.patch text/x-patch 36.2 KB

In response to

Browse pgsql-hackers by date

  From Date Subject
Previous Message Ayush Tiwari 2026-11-20 14:14:20 Re: [Patch] Batch fsyncs when recycling WAL segments