Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 6 Apr 2022 14:57:14 +0000
From:      John F Carr <jfc@mit.edu>
To:        "egoitz@ramattack.net" <egoitz@ramattack.net>
Cc:        "freebsd-fs@freebsd.org" <freebsd-fs@freebsd.org>
Subject:   Re: Desperate with 870 QVO and ZFS
Message-ID:  <D7E66B1E-C036-4362-AB69-3875C8F05E83@exchange.mit.edu>
In-Reply-To: <4e98275152e23141eae40dbe7ba5571f@ramattack.net>
References:  <4e98275152e23141eae40dbe7ba5571f@ramattack.net>

next in thread | previous in thread | raw e-mail | index | archive | help
On Apr 6, 2022, at 07:15 , egoitz@ramattack.net wrote:
>=20
> Good morning,
>=20
> I write this post with the expectation that perhaps someone could help me=
 <d8974688.gif>
>=20
> I am running some mail servers with FreeBSD and ZFS. They use 870 QVO (no=
t EVO or other Samsung SSD disks) disks as storage. They can easily have fr=
om 1500 to 2000 concurrent connections. The machines have 128GB of ram and =
the CPU is almost absolutely idle. The disk IO is normally at 30 or 40% per=
cent at most.
>=20
> The problem I'm facing is that they could be running just fine and sudden=
ly at some peak hour, the IO goes to 60 or 70% and the machine becomes extr=
emely slow. ZFS is all by default, except the sync parameter which is set d=
isabled. Apart from that the ARC is limited to 64GB. But even this is extre=
mely odd. The used ARC is near 20GB. I have seen, that meta cache in arc is=
 very near to the limit that FreeBSD automatically sets depending on the si=
ze of the ARC you set. It seems that almost all ARC is used by meta cache. =
I have seen this effect in all my mail servers with this hardware and softw=
are config.
>=20

My system with

nvd0: <Samsung SSD 970 EVO 1TB> NVMe namespace

has a problem with high write volume.  If I build llvm with debugging symbo=
ls, which writes about 70 GB, the filesystem nearly grinds to a halt.  I ha=
ve to use a spinning disk to get decent performance on this workload.  Ther=
e is some old talk on the mailing lists about certain drives handling TRIM =
commands badly.  See comment by Ted Ts'o here: https://forums.freebsd.org/t=
hreads/ssd-trim-maintenance.56951/

Unfortunately the documentation for adjusting trim settings is out of date.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?D7E66B1E-C036-4362-AB69-3875C8F05E83>