Linux NFS client and FreeBSD server strangeness

[Kaya Saman]

If I recall correctly the "sync" option is default. Though it might be
different depending on the Linux distro in use?

I use this: vers=3,defaults,auto,tcp,rsize=8192,wsize=8192

though I could get rid of the tcp as that's also a default. The rsize
and wsize options are for running Jumbo Frames ie. larger MTU then
1500; in my case 9000 for 1Gbps links.

At least for Arch Linux the guide does explain quite a bit, though the
'man' page does too ;-) https://wiki.archlinux.org/index.php/NFS

On Wed, Apr 4, 2018 at 9:14 PM, Mike Tancsa <mike at sentex.net> wrote:
> On 4/4/2018 4:04 PM, Kaya Saman wrote:
>> I think I posted a similar topic ages ago but got no response. It would
>> be interesting to see the cause for this. I have a hunch it could be
>> something to do with the nfs implementation of Linux but it is just pure
>> conjecture on my side.
>
> I know what you mean about the zfs / zil lockups. It makes a HUGE
> difference performance wise if you run with sync disabled on writes for
> the zfs dataset. Not sure I want to do that. I am going to try later
> with ZIL on some NVE devices and see I can have the best of both worlds.
>  I have individual zfs data sets for each server to write their backups
> to.  I might be able to live with the risk of running with sync disabled
> to receive backup files with.
>
> I think I may have found the setting on the LINUX client.  If I mount
> with the options
>
> -o tcp,intr,noatime,sync,vers=3
>
> ie sync
>
> I get the "smooth" traffic patterns. Its actually a bit slower than the
> with the default async which I guess is able to burst up higher.
>
> I am not sure how dangerous running async writes are on an nfs client.
> But for now, this mystery seems to be a bit better explained to me!
>
>         ---Mike
>
>
> --
> -------------------
> Mike Tancsa, tel +1 519 651 3400 x203
> Sentex Communications, mike at sentex.net
> Providing Internet services since 1994 www.sentex.net
> Cambridge, Ontario Canada