2009-11-02 04:20:11

by Richard Scobie

[permalink] [raw]
Subject: NFS breaking under heavy load intermittently

Client: Core i7, 12GB Fedora 8 2.6.26.8-57.fc8 x86_64, NFS mount
options rsize=32768,wsize=32768,intr,hard,vers=3, Gigabit connected.

Server: Core 2 Quad Xeon, 16GB Fedora 9 2.6.27.19-78.2.30.fc9.x86_64

Under some heavy work loads where the client is pulling large files and
all CPUs are maxed out, the NFS connection "disappears" - df shows no
entry for the NFS mount and some minutes later it will reappear again.

I am inclined to suspect the client, as similar workloads from other
machines to the same server have no problem.

Suspecting the onboard Realtek NIC, I replaced it with an Intel, but the
problem still occurs.

Can anyone tell me if there is likely to be a bug in these kernel
versions that is contributing to this?

Any debugging advice would be gratefully received; other than this in dmesg:

nfs: server flash not responding, still trying
nfs: server flash not responding, still trying
nfs: server flash OK
nfs: server flash OK


there is nothing in the kerenl logs.

Regards,

Richard


2009-11-02 04:34:15

by Peter Chacko

[permalink] [raw]
Subject: Re: NFS breaking under heavy load intermittently

You can change the rsize and wsize, typically to 8k to avoid
unnecessary IP -fragmentation .

On Mon, Nov 2, 2009 at 9:30 AM, Richard Scobie <[email protected]> wr=
ote:
> Client: Core i7, 12GB Fedora 8 =A02.6.26.8-57.fc8 x86_64, NFS mount o=
ptions
> rsize=3D32768,wsize=3D32768,intr,hard,vers=3D3, Gigabit connected.
>
> Server: Core 2 Quad Xeon, 16GB Fedora 9 2.6.27.19-78.2.30.fc9.x86_64
>
> Under some heavy work loads where the client is pulling large files a=
nd all
> CPUs are maxed out, the NFS connection "disappears" - df shows no ent=
ry for
> the NFS mount and some minutes later it will reappear again.
>
> I am inclined to suspect the client, as similar workloads from other
> machines to the same server have no problem.
>
> Suspecting the onboard Realtek NIC, I replaced it with an Intel, but =
the
> problem still occurs.
>
> Can anyone tell me if there is likely to be a bug in these kernel ver=
sions
> that is contributing to this?
>
> Any debugging advice would be gratefully received; other than this in=
dmesg:
>
> nfs: server flash not responding, still trying
> nfs: server flash not responding, still trying
> nfs: server flash OK
> nfs: server flash OK
>
>
> there is nothing in the kerenl logs.
>
> Regards,
>
> Richard
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs" =
in
> the body of a message to [email protected]
> More majordomo info at =A0http://vger.kernel.org/majordomo-info.html
>