[Bug 236220] ZFS vnode deadlock on zfs_mknode

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=236220

--- Comment #24 from ncrogers at gmail.com ---
(In reply to Dennis Noordsij from comment #23)
Yes, I have a few systems that have about a 6 week uptime now after applying
the patch, and many more running the same patch with uptimes less than a month
because that's about when I started upgrading more systems. It does seem to
take up to a few weeks to manifest itself but I feel like I have enough systems
running this now that it would have happened.

In my experience it doesn't seem to matter if its a small workload or a huge
workload, eventually it strikes. My guess is the problem is unique to something
strange with how PostgreSQL handles its WAL but I have no idea what that would
be. I have full_page_writes OFF, synchronous_commit OFF, and fsync ON in my
postgresql conf. Running PostgreSQL 11.2.

I will definitely report back if I witness a hang/crash. Perhaps it would be
useful for someone to open up an issue with PostgreSQL? I am less familiar with
their bug reporting process. But given that there is definitively a bug in a
feature borrowed  from ZoL that was later fixed in ZoL upstream, it seems like
this is more of a FreeBSD problem, so I am not sure if PostgreSQL devs will
have much to say other than to perhaps help pinpoint why this is only happening
to postgres users?

-- 
You are receiving this mail because:
You are the assignee for the bug.