Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 17 May 2016 12:36:27 +0200
From:      Fabian Keil <freebsd-listen@fabiankeil.de>
To:        FreeBSD Filesystems <freebsd-fs@freebsd.org>
Subject:   Re: zfs receive stalls whole system
Message-ID:  <20160517123627.699e2aa5@fabiankeil.de>
In-Reply-To: <c090ab7bbff2fffe2a49284f9be70183@ultra-secure.de>
References:  <0C2233A9-C64A-4773-ABA5-C0BCA0D037F0@ultra-secure.de> <20160517102757.135c1468@fabiankeil.de> <c090ab7bbff2fffe2a49284f9be70183@ultra-secure.de>

next in thread | previous in thread | raw e-mail | index | archive | help
--Sig_/Fulk3QySoETDWNL8l4bPeY/
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: quoted-printable

rainer@ultra-secure.de wrote:

> Am 2016-05-17 10:27, schrieb Fabian Keil:
> > Rainer Duffner <rainer@ultra-secure.de> wrote:
> >  =20
> >> I have two servers, that were running FreeBSD 10.1-AMD64 for a long=20
> >> time, one zfs-sending to the other (via zxfer). Both are NFS-servers=20
> >> and MySQL-slaves, the sender is actively used as NFS-server, the=20
> >> recipient is just a warm-standby, in case something serious happens=20
> >> and we don=E2=80=99t want to wait for a day until the restore is back =
in=20
> >> place. The MySQL-Slaves are actively used as read-only servers (at the=
=20
> >> application level, Python=E2=80=99s SQL-Alchemy does that, apparently).
> >>=20
> >> They are HP DL380G8 (one CPU, hexacore) with over 128 GB RAM (I think=
=20
> >> one has 144, the other has 192).
> >> While they were running 10.1, they used HP P420 RAID-controllers with=
=20
> >> individual 12 RAID0 volumes that I pooled into 6-disk RAIDZ2 vdevs.
> >> I use zfsnap to do hourly, daily and weekly snapshots. =20
> > [...] =20
> >> Now, when I do a zxfer, sometimes the whole system stalls while the=20
> >> data is sent over, especially if the delta is large or if something=20
> >> else is reading from the disk at the same time (backup agent).
> >>=20
> >> I had this before, on 10.0 (I believe, we didn=E2=80=99t have this in =
9.1=20
> >> either, IIRC) and it went away in 10.1. =20
> >=20
> > Do you use geli for swap device(s)? =20
>=20
>=20
> Yes, I do.
> /dev/mirror/swap.eli		none	swap	sw		0	0
>=20
> Bad idea?

It can cause deadlocks and poor performance when paging.

This was recently fixed in ElectroBSD and I intend to submit
the patch in a couple of days after a bit more stress testing.

The patch is already available at:
https://www.fabiankeil.de/sourcecode/electrobsd/GELI-Use-a-dedicated-uma-zo=
ne-for-writes.diff

Fabian

--Sig_/Fulk3QySoETDWNL8l4bPeY/
Content-Type: application/pgp-signature
Content-Description: OpenPGP digital signature

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

iEYEARECAAYFAlc69CsACgkQBYqIVf93VJ1CEgCgt1OGGtUcZ/RH421bYTH0ZXaK
GsgAn32stEPMDuiCyd5favEJTAKnJtT/
=3xI0
-----END PGP SIGNATURE-----

--Sig_/Fulk3QySoETDWNL8l4bPeY/--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20160517123627.699e2aa5>