[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 #7 from Dave Chinner (david@fromorbit.com) ---
On Mon, Jul 03, 2023 at 07:56:36PM +0000, bugzilla-daemon@kernel.org wrote:
> https://bugzilla.kernel.org/show_bug.cgi?id=217572
> 
> --- Comment #6 from Christian Theune (ct@flyingcircus.io) ---
> Daniel pointed me to this patch they're considering as a valid fix:
>
> https://lore.kernel.org/linux-fsdevel/20221129001632.GX3600936@dread.disaster.area/

No, that has nothing to do with the problem you are seeing on 6.1.31
kernels. That was a fix for a regression introduced in 6.3-rc1, and
hence does not exist in 6.1.y kernels.

The problem you are tripping over appears to be a livelock in the
page cache iterator infrastructure, not an issue with the filesystem
itself. This has been seen occasionally (maybe once every couple of
months of testing across the entire dev community) during testing
since large folios were enabled in the page cache, but nobody has
been able to reproduce it reliably enough to be able to isolate the
root cause and fix it yet.

If you can reproduce it reliably and quickly, then putting together
a recipe that we can use to trigger it would be a great help.

-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.