From: Frans Pop <elendil@planet.nl>
To: Trond Myklebust <trond.myklebust@fys.uio.no>
Cc: linux-nfs@vger.kernel.org,
Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
netdev@vger.kernel.org
Subject: Re: [2.6.30.1] Significant latency playing video file from NFS4 share
Date: Tue, 21 Jul 2009 23:54:10 +0200 [thread overview]
Message-ID: <200907212354.11174.elendil@planet.nl> (raw)
In-Reply-To: <1248199105.21343.16.camel@heimdal.trondhjem.org>
On Tuesday 21 July 2009, Trond Myklebust wrote:
> This would be where you are losing your performance. 1998.933 ms rtt
> means that you are seeing a 2 second delay from the instant the RPC
> client pushes the request into the socket to the moment that socket
> receives a reply from the server.
>
> Basically, you have either a _very_ slow server, or (more likely) a
The server should be fast enough.
> networking problem. A tcpdump ought to be able to show what the problem
> is: whether it is the packets getting ACKed very slowly, and/or if you
> have dropped packet issues or if it is truly a problem with the server
> taking a long time to reply.
I did run a traceroute alongside and that did not show packet loss, but I
do now see some TCP data loss events and retransmits of segments in
'netstat -s', and "TCP Previous segment lost" with varying time gaps in
wireshark. So I guess it is indeed a networking problem.
As I see it both with wired and wireless on the client, the problem must
be either in the server or the switch near the server (replacing the
cable from the server did not help).
I suspect the integrated e1000e NIC. Guess I'll add an extra NIC in the
server soon to see if that helps.
Thanks a lot for your help!
P.S. I see one strange thing in wireshark: NFS (or the networking stack)
seems to switch between small and large package (?) sizes occasionally in
the same stream. Is that normal?
Source Destination Protocol Info
[...]
<server> <client> RPC Continuation
<server> <client> RPC Continuation
<client> <server> TCP kink > nfs [ACK] ...
<server> <client> RPC Continuation
<server> <client> RPC Continuation
<client> <server> TCP kink > nfs [ACK] ...
<server> <client> TCP [TCP segment of reassembled PDU]
<server> <client> TCP [TCP segment of reassembled PDU]
<client> <server> TCP kink > nfs [ACK] ...
<server> <client> TCP [TCP segment of reassembled PDU]
<server> <client> TCP [TCP segment of reassembled PDU]
<client> <server> TCP kink > nfs [ACK] ...
[...]
<server> <client> TCP [TCP segment of reassembled PDU]
<server> <client> RPC Continuation (total: 65535 bytes)
<client> <server> TCP kink > nfs [ACK] ...
<server> <client> RPC Continuation
<server> <client> RPC Continuation
<client> <server> TCP kink > nfs [ACK] ...
[...]
prev parent reply other threads:[~2009-07-21 21:54 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-07-17 20:50 [2.6.30.1] Significant latency playing video file from NFS4 share Frans Pop
2009-07-17 22:51 ` Trond Myklebust
2009-07-17 23:31 ` Frans Pop
2009-07-21 17:31 ` Frans Pop
2009-07-21 17:58 ` Trond Myklebust
2009-07-21 21:54 ` Frans Pop [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200907212354.11174.elendil@planet.nl \
--to=elendil@planet.nl \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-nfs@vger.kernel.org \
--cc=netdev@vger.kernel.org \
--cc=trond.myklebust@fys.uio.no \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox