Subject: Re: NFS problem.
To: Johnny Billquist <firstname.lastname@example.org>
From: Johnny Billquist <email@example.com>
Date: 12/10/2005 01:36:25
Johnny Billquist wrote:
> Matthias Scheler wrote:
>> On Sat, Dec 10, 2005 at 12:24:41AM +0100, Johnny Billquist wrote:
>>> Wasn't there a problem in the past with large UDP packets and NFS?
>> Yes, but it hasn't changed and never will. Large UDP packets are sent
>> as IP fragments. If you lose one of the IP fragments the whole UDP
>> packet is lost because there is no selective retransmit. When a machine
>> e.g. loses 5% of incoming packets at least one of the IP fragments
>> of a 32KB UDP packet will always get lost. Retries will not help because
>> another single lost packet will prevent the reception of the UDP packet.
> The problem is that some hardware cannot send that many packets in a
> burst like NFS does. So it will never succeed.
> Ah, found it.
> NFS_RSIZE, NFS_WSIZE and NFS_READDIRSIZE
> you can set those options in your config file to change the size of
> those packets.
Bingo! That was it. Stupid of me not to remember, since I have had
similar problems on other machines in the past.
But I believe it was only i386 which used to have 32K reads and writes
over NFS before. Looking at sys/nfs/nfs.h it appears the default is 8K.
I don't know which it actually defaults to, but changing it to 2K made
my VAX happy.
Now we just have to figure out the build failure in lib/libc/rpc/svc_vc.c
Has anyone looked at it (apart from Tom Ivar Helbekkmo)?
Johnny Billquist || "I'm on a bus
|| on a psychedelic trip
email: firstname.lastname@example.org || Reading murder books
pdp is alive! || tryin' to stay hip" - B. Idol