[Bug 236220] ZFS vnode deadlock on zfs_mknode

[bugzilla-noreply at freebsd.org]

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

Julien Cigar <julien at perdition.city> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |julien at perdition.city

--- Comment #14 from Julien Cigar <julien at perdition.city> ---
I'm also experimenting something similar on 12.0-RELEASE + ZFS, also running
PostgreSQL 11 (in jail, with nullfs mounts).

I've seen this on two machines.

The first one is a small HPE Proliant Microserver Gen 8 with 16GB RAM. It
started when I upgraded from 10.3 to 12.0, so I don't know if 11.x is affected
too. As OP, sometimes I can remote SSH, sometimes not, and a simple ls almost
always hangs. A shutdown -r now hangs too after syncing buffers. For some
unknow reason the problem didn't occured again (3 months now) since I forced
ARC_MAX to 8GB (vfs.zfs.arc_max=8589934592).

The second one is an HPE Proliant DL160Gen9 with 64GB of RAM (freshly installed
12.0-RELEASE). Symptoms are the same as above, with ARC_MAX set to 24GB
(vfs.zfs.arc_max=25769803776).

(I'll try to procstat -kk -a next time)

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