[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 #4 from Christian Theune (ct@flyingcircus.io) ---
I've only seen this once, so no indication on older or newer kernels, yet,
either with smaller or larger databases. I fortunately could repair the
PostgreSQL database with a FULL VACUUM on the table and then the dump worked
fine again.

Hanging in the past typically indicated a network storage issue, so I'm aware
of the multiple causes the hung tasks can have, I still appreciate the link. :)

At the time of the hung tasks, I can see almost no IO (but also no IO pressure)
and 60% (of 3 CPUs) are reported as using up SYSTEM time. 

Something that made me think XFS was that we ended up with inconsistent data
within PostgreSQL which I haven't seen in a decade.

Nevertheless, it appears this might be a MM issue as I stumbled over this
inquiry which also mentions a 6.1 kernel:
https://www.spinics.net/lists/kernel/msg4783004.html

I'm trying to get in touch with Daniel to see whether his analysis went
anywhere ...

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

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