[ovs-discuss] Port satturation?

Bryan Osoro bosoro at gmail.com
Wed Oct 12 01:06:09 UTC 2011


This is a very common situation in the physical world, simply known as
unknown unicast flooding and is a direct result of what Jesse just mentioned
(i.e. the ARP cache being longer than the MAC address table timer).  There
are also other situations that can cause it, but fundamentally, if VMA knows
the IP/MAC of VMB but it hasn't sent a packet in over 60 seconds, OVS will
have aged out the MAC and will not re-learn it until you have bi-directional
traffic from said host.

Here's an example of other causes, to demonstrate that this is not a
phenomena of OVS
http://www.cisco.com/en/US/products/hw/switches/ps700/products_tech_note09186a00801d0808.shtml



-bryan


2011/10/11 Jesse Gross <jesse at nicira.com>

> 2011/10/11 Vladimir Nikolić <vladimir.nikolic at amis.si>:
> > The switch should learn syslog server mac address from the arp response,
> > and send traffic only to uplink.
>
> It learns it and then it times out, just as I said in my first response.
>
> The problem is that your ARP cache timeout is longer than the switch's
> MAC learning timeout and there are no other transmissions.
>
> Open vSwitch uses a 60 second MAC table timeout, many physical
> switches use 300 seconds but in both cases it is shorter than the 1200
> seconds that you are using for your ARP cache.  That's why you see
> this behavior.
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20111011/823a504d/attachment.html>


More information about the discuss mailing list