Re: Spurious "apparent wraparound" via SimpleLruTruncate() rounding

Noah Misch <noah@leadboat.com>

From: Noah Misch <noah@leadboat.com>

To: pgsql-hackers@postgresql.org

Date: 2019-02-17T04:09:13Z

Lists: pgsql-hackers

Attachments

slru-truncate-modulo-v2.patch (text/plain) patch v2

On Wed, Feb 13, 2019 at 11:26:23PM -0800, Noah Misch wrote:
> On further study, I was able to reproduce data loss

> Fixes are available:
> 
> a. Fix the rounding in SimpleLruTruncate().  (The patch I posted upthread is
>    wrong; I will correct it in a separate message.)

Here's a corrected version.  I now delete a segment only if both its first
page and its last page are considered to precede the cutoff; see the new
comment at SlruMayDeleteSegment().

Commits

Prevent excess SimpleLruTruncate() deletion.
- c424c75f5382 9.5.25 landed
- 9afba9b0e282 10.16 landed
- 677f6cb1d987 11.11 landed
- 66087f7e922c 12.6 landed
- 1a31d8c52db4 9.6.21 landed
- 6eb3fc7fcd89 13.2 landed
- 6db992833c04 14.0 landed
Fix unlinking of SLRU segments.
- c732c3f8c122 14.0 cited
Defer flushing of SLRU files.
- dee663f78439 14.0 cited
Change XID and mxact limits to warn at 40M and stop at 3M.
- cd5e82256de5 14.0 cited