Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 14 Oct 2005 15:54:59 -0400
From:      Chuck Lever <cel@citi.umich.edu>
To:        rick@snowhite.cis.uoguelph.ca
Cc:        fs@freebsd.org
Subject:   Re: FreeBSD NFS server not responding to TCP SYN packets from	Linux/SunOS clients
Message-ID:  <43500D13.1030702@citi.umich.edu>
In-Reply-To: <200510141742.NAA23853@snowhite.cis.uoguelph.ca>
References:  <200510141742.NAA23853@snowhite.cis.uoguelph.ca>

next in thread | previous in thread | raw e-mail | index | archive | help
This is a multi-part message in MIME format.
--------------070508070106000201080603
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit

rick@snowhite.cis.uoguelph.ca wrote:
>>Tcp always throws away retransmissions.  Doesn't matter whether the data is
>>still in the receive socket queue or not.  The nfs server will never see
>>replayed requests as a result of tcp retransmission.
> 
> 
> 1) Yes, of course. (Brain fart on my part. I mentioned I wasn't a TCP wizzard,
> didn't I?:-)
> There is still the problem that not all
> clients obey the "only retry an RPC after a disconnect/reconnect" rule,

where is that rule stated?  most NFS clients i am aware of retransmit an 
RPC after 60 seconds over TCP.

> but that should be fixed for NFSv4.

agreed.

>>  The problem is "how do you make sure the nfsd threads don't start a
>>  request if the disk I/O subsystem is backlogged".
>>
>>Isn't this simply a matter of choosing the right number of nfsd threads?
> 
> 
> That's the only mechanism available to-day (at least for my server and,
> I think, the current FreeBSD one). Unfortunately load is pretty dynamic
> and it might be nice if "the number of active nfsd threads" changed to
> reflect that, without manual sysadmin intervention.
> 
> However, given 1) and clients that only retry RPCs after a reconnect, I
> don't think it will be all that critical. The clients will see the slow
> response to RPCs and new requests will be limited by the number of threads
> doing I/O on the client.

that assumes the client uses a threaded implementation too, doesn't it?

--------------070508070106000201080603--



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43500D13.1030702>