[Bug 217572] Initial blocked tasks causing deterioration over hours until (nearly) complete system lockup and data loss with PostgreSQL 13

[bugzilla-daemon@kernel.org]

https://bugzilla.kernel.org/show_bug.cgi?id=217572

--- Comment #12 from Dave Chinner (david@fromorbit.com) ---
On Thu, Sep 28, 2023 at 12:39:05PM +0000, bugzilla-daemon@kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=217572
> 
> --- Comment #10 from Christian Theune (ct@flyingcircus.io) ---
> @Dave I've seen this a second time now today. It also ended up being a copy
> statement with postgresql and I'm going to take a consistent snapshot of the
> full disk right now.

Have you tried seeing if commit cbc02854331e ("XArray: Do not return
sibling entries from xa_load()") fixes the problem? That has been
marked for stable, so I'd expect that it eventually makes it back to
6.1.y or whatever LTS kernel you are running now....

As for debug, this isn't an XFS problem - it's a page cache problem
when high-order folios are in use (which XFS uses) - so you need to
be looking at in-memory page cache state that is being tripped over
first....

-Dave.

-- 
You may reply to this email to add a comment.

You are receiving this mail because:
You are watching the assignee of the bug.