This is an archived copy of the Xen.org mailing list, which we have preserved to ensure that existing links to archives are not broken. The live archive, which contains the latest emails, can be found at http://lists.xen.org/
Home Products Support Community News


Re: [Xen-users] high ? number of dropped packet on xen vif

To: xen-users@xxxxxxxxxxxxxxxxxxx
Subject: Re: [Xen-users] high ? number of dropped packet on xen vif
From: Matthieu PATOU <matxen@xxxxxxxxx>
Date: Tue, 27 Dec 2005 23:21:06 +0100
Delivery-date: Tue, 27 Dec 2005 22:25:11 +0000
Envelope-to: www-data@xxxxxxxxxxxxxxxxxxx
In-reply-to: <A95E2296287EAD4EB592B5DEEFCE0E9D409CA2@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
List-help: <mailto:xen-users-request@lists.xensource.com?subject=help>
List-id: Xen user discussion <xen-users.lists.xensource.com>
List-post: <mailto:xen-users@lists.xensource.com>
List-subscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=subscribe>
List-unsubscribe: <http://lists.xensource.com/cgi-bin/mailman/listinfo/xen-users>, <mailto:xen-users-request@lists.xensource.com?subject=unsubscribe>
References: <A95E2296287EAD4EB592B5DEEFCE0E9D409CA2@xxxxxxxxxxxxxxxxxxxxxxxxxxx>
Sender: xen-users-bounces@xxxxxxxxxxxxxxxxxxx
User-agent: Mozilla Thunderbird 1.0.7 (X11/20050923)
Hello i did some test with xen 2.0.6 and xen 3.0.0
The test done with xen 2.0.6 is switching from UDP to TCP and it works
great the transfert speed has increased.
And of course, with the nature of TCP (in term of network sessions)
there is no more timeout and hangs.
I've also installed a box with  ubuntu 5.10 as Dom0 and Debian sarge as
On this box with a bridged configuration there is no communication between the client (a standalone linux box running ubuntu 5.04) and the server when i try to copy a big file (1 Mo or more). It appears that fragmented packet (or at least packet with 5 or more fragment) didn't cross the bridge and so the server and so the client keeps on trying to send the first bytes of the file and fail because there is no return not even a failure but small file are OK (if the try is done before the test of the big file). I've also done a test with a different setup on the Xen 3.0 box, i've setup dom0 as a router between the domU serving nfs file and the client. In this configuration i haven't got hang and the time needed to copy the file is less than the original configuration (xen 2.0.6 + bridging) but when i start ethereal i still see some packet retransmission and i think it's not normal because the client and xen server are on the same switch with good cable and only a couple of meter away. Any idea of why there is this huge problem with the bridge and why there is still packet loss (or not acknoledge quickly enougth) ?

Hello all,
it seems i'm facing a  problem with xen networking.
I have some hang up with my NFS server running in domU.
I noticed that when i got a hang up i also see an increase of the number of dropped packet of the vif interface.

Which version of Xen are you running here, 3.0.0?

Is this a soft hang that later recovers?

it seems that dom0 has some trouble with sending the packet to the domU Currently the rate of dropped packet (the TX ones ) is 10% of the total transmited.

Which counters are you referring to here? The TX fields of the vifX.0

I try to increase dom0 memory but i doesn't seems to have an effect on the rate of dropped packet.
Is there some one who has also faced this problem ?
Do you have an idea of what can i try to solve the problem ?

If you built from source, please modify xen/Rules.mk to turn on debug=y
and then rebuild.



Xen-users mailing list

<Prev in Thread] Current Thread [Next in Thread>