Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 24 Jun 2018 12:03:29 +0200
From:      Alexander Leidinger <Alexander@leidinger.net>
To:        current@freebsd.org
Subject:   numa involved in instability and swap usage despite RAM free?
Message-ID:  <20180624120329.Horde.HWORumQ7Ng1KAUeviJNtoc3@webmail.leidinger.net>

next in thread | raw e-mail | index | archive | help
This message is in MIME format and has been PGP signed.

--=_-oBR2oHEOll7in8gWqzLcvA
Content-Type: text/plain; charset=utf-8; format=flowed; DelSp=Yes
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

Hi,

I don't have hard evidence, but there is enough "smell" to open up a=20=20
discussion...

Short:
Can=20it be that enabling numa in the kernel is the reason why some=20=20
people=20see instability with zfs and usage of swap while a lot of free=20=
=20
RAM=20is available?

Long:
I have a dual-socket Xeon system (E5620 + L5630... yes, not the same,=20=20
but=20compatible enough to be able to run together) with 64 GB RAM. I=20=20
run=20-current on it (currently it's at r333966 and it was for all the=20=
=20
tests=20below).

What I see with numa enabled and no zfs patches is, that at some point=20=
=20
I=20have half the RAM free, swap is in use, and after a lot of compiling=20=
=20
ports=20in different jails ZFS comes to a halt (sometimes I can unblock=20=
=20
by=20killing a compile, sometimes I can't even kill, only way out is=20=20
power-cycle).=20I've seen this around twice a week.

When I keep numa enabled and have applied this ZFS patch=20=20
https://reviews.freebsd.org/D7538=20the bahavior changes. AFter a while=20=
=20
half=20of the RAM is free, swap is in use, and after enough compiling=20=20
ports=20in jails I get a panic (unfortunately not enough debug info in=20=
=20
the=20textdump to know exactly what he problem is).

Since 2 weeks I have numa compiled out of the kernel (and still the=20=20
ZFS=20patch inside). The system is down to 17 GB free and NO swap in=20=20
use.=20I'm compiling ports in 16 jails (one of them with parts of KDE5 =3D=
=20=20
currently=20about 700 ports compiled) and not a single issue like the=20=20
above.

For=20everyone with swap issues or ZFS issues similar to the ones I=20=20
see...=20do you have numa enabled and can you please try without and=20=20
report=20back?

Can it be that if memory request can not be fulfilled from one numa=20=20
domain,=20there is no fallback to another numa domain for all the=20=20
various=20kinds of memory allocation we have in the kernel=20=20
(contigmem/no-sleep/...)?

Bye,
Alexander.

--=20
http://www.Leidinger.net=20Alexander@Leidinger.net: PGP 0x8F31830F9F2772BF
http://www.FreeBSD.org    netchild@FreeBSD.org  : PGP 0x8F31830F9F2772BF

--=_-oBR2oHEOll7in8gWqzLcvA
Content-Type: application/pgp-signature
Content-Description: Digitale PGP-Signatur
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQIcBAABAgAGBQJbL2xxAAoJEKrxQhqFIICEVcsP/jVTtyEtb92DPBitIacgUk1F
3xzRc+0SfWYqyU8PAHsIRkJQFdvQ7rvLqwNznqvkLdi3FvmT+6PKJcORhHrNBMEo
KhJGByGFPAdpnAoA7ZVJY7r6RWJsNU61AvPdgn0ZORPX73cbuo5zeCfwpIiYPGEH
oja29zNA7XLnh9AZ5NJMl3zmvCEUOOONkENoUE/tDi4YMVy1m8EVmtSFAsdtBYHU
G9incSchNzRx3tkK8kojknLjZCFQn44yOKYhANKwgMSECR5sZuvAcQaSLUNUvdcr
6OpjCFcHWsTFIr5vJvEqml+rSSnvt89q5JZBsP9T8/6dXs1rUuBKKLkVWR2A/bT9
d7c5WX4xFlk6vRRfcZ6pQ6GQQOs06IaxxPvRXDAdJcFCMj1gKatKumg55+IXxbwQ
ovxuHay+YfmQxi/DuO0HAeh5JNd1NL3u2Dpb9Xs+xey0yKQXorFTU4ADiqcWXp/v
kAiz0Cyf3gnaYuRA6bRZdUdTThj/3NDKKWztcX6pe+VGepSBBRsskEcfIns6oNgG
n+CnkUIp5sjos+kG0iXD7d+ZqVPiHsdiWdvLzOuJD1bjagDbMD23eoA+m3uYLcKn
buf/DJoW8VthcvlKPsh9Lxh2ChfiT3C1MVS0BTGeN6tfOWsPpg9E5c0e1mouWV9+
nMgw+XE0WfA2UDrP0X8k
=ckNr
-----END PGP SIGNATURE-----

--=_-oBR2oHEOll7in8gWqzLcvA--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180624120329.Horde.HWORumQ7Ng1KAUeviJNtoc3>