[ovs-discuss] OVS in-band mode: working but need help understanding the reason

Ben Pfaff blp at nicira.com
Thu May 22 16:31:19 UTC 2014


On Thu, May 22, 2014 at 05:53:56PM +0200, Dani Camps wrote:
> I think the problem I am having is that the "ARP request to controller"
> hidden rule is not matching. This is what happens:
> 
> 1) In the switch (192.168.56.203) connected to the controller
> (192.168.56.103) I see ARP Requests coming from the other switch:
> 
> # tcpdump -i s3-eth0
> 19:47:32.818889 ARP, Request who-has 192.168.56.103 tell 192.168.56.202,
> length 28
> 19:47:33.816421 ARP, Request who-has 192.168.56.103 tell 192.168.56.202,
> length 28
> ...
> 
> 2) However when I dump the status of the hidden flows in the switch
> connected to the controller I get this:
> 
> duration=485s, n_packets=851, n_bytes=924434,
> priority=180007,tcp,nw_dst=192.168.56.103,tp_dst=6633,actions=NORMAL
> duration=485s, n_packets=1, n_bytes=60,
> priority=180006,arp,arp_spa=192.168.56.103,arp_op=1,actions=NORMAL
> duration=485s, n_packets=0, n_bytes=0,
> priority=180005,arp,arp_tpa=192.168.56.103,arp_op=2,actions=NORMAL
> duration=485s, n_packets=860, n_bytes=63226,
> priority=180008,tcp,nw_src=192.168.56.103,tp_src=6633,actions=NORMAL
> duration=485s, n_packets=1, n_bytes=60,
> priority=180001,arp,dl_dst=f6:b8:41:3c:d6:45,arp_op=2,actions=NORMAL
> duration=478s, n_packets=0, n_bytes=0,
> priority=180003,arp,dl_dst=08:00:27:a9:08:16,arp_op=2,actions=NORMAL
> duration=485s, n_packets=0, n_bytes=0,
> priority=180000,udp,in_port=LOCAL,dl_src=f6:b8:41:3c:d6:45,tp_src=68,tp_dst=67,actions=NORMAL
> duration=485s, n_packets=1, n_bytes=42,
> priority=180002,arp,dl_src=f6:b8:41:3c:d6:45,arp_op=1,actions=NORMAL
> duration=478s, n_packets=0, n_bytes=0,
> priority=180004,arp,dl_src=08:00:27:a9:08:16,arp_op=1,actions=NORMAL
> table_id=254, duration=486s, n_packets=0, n_bytes=0,
> priority=0,reg0=0x3,actions=drop
> table_id=254, duration=486s, n_packets=320, n_bytes=30696,
> priority=0,reg0=0x1,actions=controller(reason=no_match)
> table_id=254, duration=486s, n_packets=0, n_bytes=0,
> priority=0,reg0=0x2,actions=drop
> 
> 
> Where specially the following rule is interesting:
> 
> duration=485s, n_packets=0, n_bytes=0,
> priority=180005,arp,arp_tpa=192.168.56.103,arp_op=2,actions=NORMAL
> 
> I understand that this rule is saying that there are no matches on ARP
> Request to the controller from other switches. But this is wrong because we
> are seeing the ARP Request on the interface. As I said in my previous
> email, eventually something changes and the rules starts to match.
> 
> Is this behavior correct?

Without thinking about what you specifically are pointing at, normally
it should only take a few seconds to connect to a controller, even
through a second Open vSwitch running in in-band mode.

Now, one thing I've noticed over the last few weeks is that I've been
having some trouble with the in-band setup in my own set of test VMs.
I haven't had a chance to fully investigate, but along with your
report, I wonder whether there's been some regression in the in-band
control implementation.

What version of Open vSwitch are you using?  (I normally test the
latest from the master branch.)



More information about the discuss mailing list