Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 17 Oct 2007 22:12:35 +0200
From:      Pawel Jakub Dawidek <pjd@FreeBSD.org>
To:        Andriy Gapon <avg@icyb.net.ua>
Cc:        freebsd-geom@freebsd.org
Subject:   Re: gjournal: FLUSHCACHE timed out
Message-ID:  <20071017201235.GD50219@garage.freebsd.pl>
In-Reply-To: <47161FD1.5010501@icyb.net.ua>
References:  <4715C3D1.3070308@icyb.net.ua> <4715F4EE.9020104@freebsd.org> <47161FD1.5010501@icyb.net.ua>

next in thread | previous in thread | raw e-mail | index | archive | help

--zS7rBR6csb6tI2e1
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Wed, Oct 17, 2007 at 05:44:33PM +0300, Andriy Gapon wrote:
> on 17/10/2007 14:41 Eric Anderson said the following:
> > Andriy Gapon wrote:
> >> ad4s1ge (please don't pay attention to its slightly unusual name, this
> >> is for historic reasons) is a journal partition/consumer for my /var
> >> filesystem/partition/provider.
> >> Size of /var is 16G, size of the journal is slightly less than 1G (1G -
> >> 32 sectors actually). /var is UFS2 with softupdates enabled.
> >=20
> >=20
> > Pawel, correct me if I'm wrong here - but I think you really need to=20
> > turn *off* softupdates on gjournaled file systems.
>=20
> I was under a big mis-impression that I have to have softupdates enabled
> for snapshots to work. Now that I know that I was wrong I will turn off
> the softupdates. But it seems that there is nothing that would preclude
> _in principle_ combination of softupdates/gjournal. Anyway, I care only
> out of curiosity.

It's not that it won't work together, but it's just hurts performance
and memory consumption.

> >> I noticed that I get these messages only when I run 'dump' on any of my
> >> filesystems. I think that dump is using /tmp or /var/tmp for some
> >> temporary data and in my setup both of those are in /var filesystem.
> >>
> >> So my I guess is that /var is being written "too" actively and I have =
to
> >> tune some parameters to make things smooth.
> >=20
> > A few things to note:
> >=20
> > - you can turn on 'async' option for your gjournaled file system, and=
=20
> > get better performance
>=20
> will do
>=20
> > - you might be able to at the 'noatime' option to your file system moun=
t=20
> > also
>=20
> probably will do as well
>=20
> > - You might try turning your journal switch time from 10 down to 5, and=
=20
> > see if it alleviates some pressure on your disk.
>=20
> I already did this and it helped! I don't see the messages anymore.
> Thank you!
> I will try to set this back to 10 after I do away with softupdates and
> see what happens.
>=20
> Thank you very much again.

You should also try this patch:

	http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/dev/ata/ata-disk.c.diff?r1=
=3D1.201;r2=3D1.202

BIO_FLUSH timeout was way too small.

--=20
Pawel Jakub Dawidek                       http://www.wheel.pl
pjd@FreeBSD.org                           http://www.FreeBSD.org
FreeBSD committer                         Am I Evil? Yes, I Am!

--zS7rBR6csb6tI2e1
Content-Type: application/pgp-signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.4 (FreeBSD)

iD8DBQFHFmyzForvXbEpPzQRAkScAJ9rnB4eERnKYOERIFHI2mKA+1zrIgCbBmDr
UqxNSAjWPjHzGqeL8p+ewsE=
=GbMN
-----END PGP SIGNATURE-----

--zS7rBR6csb6tI2e1--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20071017201235.GD50219>