[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 #23 from Christian Theune (ct@flyingcircus.io) ---
I'm trying to come up with a reproducer and reviewed our machines. Currently
all the affected machines are running one of our database configurations (MySQL
or PostgreSQL) and one difference is that they are running with increased
readahead (128KiB default on our fleet is being increased to 1 MiB when running
MySQL or PostgreSQL).

I also reviewed the RedHat bug and all the linked discussions. The RedHat bug
smells like there might be a reproducer as the kernel dumps show uptimes of
less than 30 minutes. Also, there's another database system involved as I've
seen RocksDB (and compactions) mentioned there.

Using the PostgreSQL benchmark did not trigger anything for me in less than 1
hour even under memory pressure. I'm going to check out the rocksdb benchmark
now.

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

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