Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 22 Jan 2016 14:05:49 +0100
From:      Fabian Keil <freebsd-listen@fabiankeil.de>
To:        Mason Loring Bliss <mason@blisses.org>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: ZFS performance help sought
Message-ID:  <20160122140549.053fea0f@fabiankeil.de>
In-Reply-To: <20160122000523.GL4538@blisses.org>
References:  <20160121205139.GG4538@blisses.org> <ea3f05f9a8c20bd62a1c391b432dafe2@dweimer.net> <20160121232810.GJ4538@blisses.org> <20160121235545.GK4538@blisses.org> <20160122000523.GL4538@blisses.org>

next in thread | previous in thread | raw e-mail | index | archive | help
--Sig_/JL2.QyCqkFPVF6wmbAVcHVL
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: quoted-printable

Mason Loring Bliss <mason@blisses.org> wrote:

> On Thu, Jan 21, 2016 at 06:55:45PM -0500, Mason Loring Bliss wrote:
>=20
> > Here's what I set:
> >=20
> > vfs.zfs.arc_max=3D"4096M"
> > vfs.zfs.arc_min=3D"1024M"
> > vfs.zfs.txg.timeout=3D"3"
> > vfs.zfs.write_limit_override=3D"512M"
> >=20
> > I'm not seeing any obvious way to verify the write_limit_override setti=
ng -
> > it appears not to show up in sysctl output.
> >=20
> > I'll wait for the current big transfer to finish and then I'll try it w=
ith
> > prefetch disabled too. =20
>=20
> Disabling prefetch doesn't do a thing here - the system is still painfully
> overloaded, doing something that was simply unproblematic under Linux.
>=20
> I'd be grateful for further debugging or tuning tips. Is it possible this=
 has
> nothing to do with ZFS and that I need to play with FreeBSD's scheduling
> somehow?
>=20
> Again, FreeBSD 10.2, ZFS tuned at noted above, and with prefetch disabled.
> Eight gigs of RAM, pools less than 1TB. Doing a send/receive between pool=
s on
> different disks is bringing the system to its knees, where the literal sa=
me
> hardware under Linux/ZoL doesn't break a sweat.
>=20
> What else can I try?

It's conceivable that you are seeing the result of lock contention that
is slowing down arc_get_data_buf(), for details see:
https://www.fabiankeil.de/gehacktes/electrobsd/zfs-arc-tuning/
You could try the referenced DTrace script to check if that's the case.

Even if it's a different issue, you may be able to work around it by
throttling the send/receive throughput with mbuffer or a similar tool.

Fabian

--Sig_/JL2.QyCqkFPVF6wmbAVcHVL
Content-Type: application/pgp-signature
Content-Description: OpenPGP digital signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2

iEYEARECAAYFAlaiKS8ACgkQBYqIVf93VJ25xgCfU/7/ZXZiZeSgvPonL+2ob4XH
1C0AoLK5e5x1CTi3JWsx4217yzHiFQ3v
=Dmk9
-----END PGP SIGNATURE-----

--Sig_/JL2.QyCqkFPVF6wmbAVcHVL--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160122140549.053fea0f>