[ovs-discuss] Flood test with xen/openvswitch

Sébastien Riccio sr at swisscenter.com
Wed Sep 7 18:53:14 UTC 2011

On 07.09.2011 17:12, Ben Pfaff wrote:
> On Wed, Sep 07, 2011 at 01:28:16PM +0200, S?bastien Riccio wrote:
>> I just did a test to see how openvswitch handle a flood from a
>> virtual machine on a xen
>> host using it as the networking layer.
>> I just issued a :
>> vm1# hping3  -S -L 0 -p 80 -i u100
>> options I used are:
>> -S set SYN tcp flag
>> -L set ACK tcp flag
>> -p destination port
>> -i u100 = interval between packets in micro seconds
>> This results in a cpu usage up to 97% by the ovs-vswitchd process in
>> the dom0.
>> Letting it go for a few minutes turns the whole xen host
>> unresponsive to network access
>> and must then be accessed from the local console.
> When you report a bug, it is important to give all the important
> details.  It is especially important for a performance-related bug like
> this.
> Please give us the following information:
>          * The Open vSwitch version number (as output by "ovs-vswitchd
>            --version").
>          * The Git commit number (as output by "git rev-parse HEAD"),
>            if you built from a Git snapshot.
>          * Any local patches or changes you have applied (if any).
>          * The kernel version on which Open vSwitch is running (from
>            /proc/version) and the distribution and version number of
>            your OS (e.g. "Centos 5.0").
>          * The contents of the vswitchd configuration database (usually
>            /etc/openvswitch/conf.db).
>          * The output of "ovs-dpctl show".
>          * If you have Open vSwitch configured to connect to an
>            OpenFlow controller, the output of "ovs-ofctl show<bridge>"
>            for each<bridge>  configured in the vswitchd configuration
>            database.

Hi Ben,

Sorry, cross posting was unintentional. I wasn't sure where was the best 
place to post this, so
I first sent it to the openvswitch list but then sent it the xen list 
too and forgot to  mention it.

When I replied to you I've noticed the addresses were proxied by 
gmane.org or something
like this and it doesn't seem to have made it to the list, so I'm 
replying here again.

For the details about the versions:

root at xen-blade13:~# ovs-vswitchd --version
ovs-vswitchd (Open vSwitch) 1.2.1+build0
Compiled Sep  6 2011 01:01:15
OpenFlow versions 0x1:0x1

It's the one from de debian unstable repository.

The kernel is Linux version 3.0.0-scxen-amd64 (root at xen-blade13) (gcc 
version 4.6.1 (Debian 4.6.1-4) ) #2 SMP Fri Aug 5 08:12:00 CEST 2011
It's the 3.0.0 stock kernel with debian .config recompiled with a vga 
fix patch for xen on running on debian unstable

root at xen-blade13:~# ovs-dpctl show
system at xapi0:
         lookups: frags:0, hit:542619, missed:83504, lost:10
         port 0: xapi0 (internal)
         port 1: eth1
         port 2: eth0
         port 3: bond0 (internal)
system at xapi1:
         lookups: frags:0, hit:1973235, missed:711045, lost:1416
         port 0: xapi1 (internal)
         port 1: xapi5 (internal)
         port 2: xapi2 (internal)
         port 3: xapi4 (internal)
         port 4: eth2
         port 5: eth3
         port 6: bond1 (internal)
         port 7: vif1.0
         port 26: vif6.0
         port 27: vif7.0

The conf.db looks quite huge, I don't know why. I've attached it to this 
reply zipped.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: conf.db.gz
Type: application/gzip
Size: 74393 bytes
Desc: not available
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20110907/eec42e10/attachment-0002.bin>

More information about the discuss mailing list