Date: Sat, 11 Apr 2020 17:36:37 +0000 (UTC) From: Chris Ross <cross+freebsd@distal.com> To: freebsd-fs <freebsd-fs@freebsd.org>, freebsd-stable@freebsd.org Subject: ZFS server has gone crazy slow Message-ID: <2182C27C-A5D3-41BF-9CE9-7C6883E43074@distal.com>
next in thread | raw e-mail | index | archive | help
I have a FreeBSD 11.3-STABLE server that is my router, using a ZFS mirror (=
of two GPT disks) as it=E2=80=99s disk. It=E2=80=99s many years old, and h=
as only been misbehaving like this for a day or so. I=E2=80=99m trying to =
figure out what=E2=80=99s wrong.
I confirmed that internet connectivity isn=E2=80=99t the problem, and a reb=
oot didn=E2=80=99t fix it. (The reboot took 10-15 minutes to finish going =
multi-user, starting daemons, due to the underlying problem described below=
.)
Truss=E2=80=99ing a very basic command (date), I can see that close() and e=
xit() calls are taking 1-2 seconds. All of the files being opened are on Z=
FS, but I don=E2=80=99t know if that=E2=80=99s for sure related. Similarly=
, using shell builtin =E2=80=9Cecho foo=E2=80=9D always is immediate, but =
=E2=80=9C/bin/echo=E2=80=9D sometimes works quickly, but sometimes the clos=
e() on /var/run/ld-elf.so.hints takes 3-5 seconds.
I _think_ this is a filesystem problem. It=E2=80=99s very hard to diagnose=
because logging in, and doing anything, takes many seconds per command. z=
pool status shows my mirror as online, so I=E2=80=99m not sure where I shou=
ld check.
I=E2=80=99d appreciate any help! Thanks much=E2=80=A6
- Chris
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?2182C27C-A5D3-41BF-9CE9-7C6883E43074>