[ovs-discuss] unknown interface report by openflowd

Justin Pettit jpettit at nicira.com
Thu Jul 15 16:30:14 UTC 2010


Did you introduce a loop?  Are eth0 and eth1 both connected to the same physical switch and not segmented with something like VLANs?  That would explain the flood of traffic on your physical switch and the eventual disconnection to the controller.

--Justin


On Jul 15, 2010, at 6:24 AM, Pásztor Lénárd Zoltán wrote:

> I just downloaded the head from git repo and openflowd starts like expected.
> 
> My test case was:
> 
> 1. start a controller on host A
> ovs-controller ptcp:6633:<controller listen address>
> It started with no output, but listens on specified ip and port, so I considered it probably works.
> 
> 2. configure host B
> I connectig via ssh to host B, which has two physical interfaces: eth0, eth1.
> 
> ovs-dpctl add-dp hostb-sw0
> ovs-dpctl add-if hostb-sw0 eth1 (i connected using eth0)
> 
> I'mb using in-band way, so next is the ip config:
> ifconfig hostb-sw0 <hostb openfow switch address>
> 
> ovs-openflowd hostb-sw0 tcp:<controller listen address>
> The openflowd reports it connected with no error.
> 
> After that I disconnect my ssh session, and connect using the hostb-sw0's IP address, I can connect successfully.
> 
> Then I configured the eth0 to use as datapath for the openflow switch.
> 
> ifconfig eth0 0.0.0.0 up
> ovs-dpctl add-if hostb-sw0 eth0
> 
> ovs-dpctl show
> system at dp0:
>         flows: cur:15, soft-max:512, hard-max:262144
>         ports: cur:3, max:1024
>         groups: max:16
>         lookups: frags:0, hit:2516, missed:1598, lost:726
>         queues: max-miss:100, max-action:100
>         port 0: vm1-sw0 (internal)
>         port 1: eth0
>         port 2: eth1
> 
> And it works fine.
> 
> When I do this operation at hostC I lost the communication to hostB and hostC, and the controller write this log message:
> 
> Jul 15 13:04:19|00001|timeval|WARN|45 ms poll interval (0 ms user, 30 ms system) is over 72 times the weighted mean interval 1 ms (237 samples)
> Jul 15 13:04:19|00002|timeval|WARN|context switches: 0 voluntary, 1 involuntary
> Jul 15 13:04:19|00003|coverage|INFO|Event coverage (epoch 237/entire run), hash=22c700f9:
> Jul 15 13:04:19|00004|coverage|INFO|flow_extract                 2 /       252
> Jul 15 13:04:19|00005|coverage|INFO|poll_fd_wait                 5 /      1014
> Jul 15 13:04:19|00006|coverage|INFO|rconn_queued                 2 /       256
> Jul 15 13:04:19|00007|coverage|INFO|rconn_sent                   2 /       256
> Jul 15 13:04:19|00008|coverage|INFO|util_xalloc                 21 /      2115
> Jul 15 13:04:19|00009|coverage|INFO|vconn_received               2 /       257
> Jul 15 13:04:19|00010|coverage|INFO|vconn_sent                   2 /       258
> Jul 15 13:04:19|00011|coverage|INFO|hmap_expand                  0 /         1
> Jul 15 13:04:19|00012|coverage|INFO|mac_learning_expired         0 /         6
> Jul 15 13:04:19|00013|coverage|INFO|mac_learning_learned         0 /        18
> Jul 15 13:04:19|00014|coverage|INFO|pstream_open                 0 /         1
> Jul 15 13:04:19|00015|coverage|INFO|vconn_open                   0 /         1
> Jul 15 13:04:19|00016|coverage|INFO|82 events never hit
> Jul 15 13:06:00|00017|timeval|WARN|11 ms poll interval (0 ms user, 10 ms system) is over 17 times the weighted mean interval 1 ms (499 samples)
> Jul 15 13:06:00|00018|timeval|WARN|context switches: 0 voluntary, 1 involuntary
> Jul 15 13:06:00|00019|coverage|INFO|Event coverage (epoch 499/entire run), hash=57205ddd:
> Jul 15 13:06:00|00020|coverage|INFO|flow_extract               100 /       741
> Jul 15 13:06:00|00021|coverage|INFO|poll_fd_wait                 6 /      2418
> Jul 15 13:06:00|00022|coverage|INFO|rconn_queued               100 /       747
> Jul 15 13:06:00|00023|coverage|INFO|rconn_sent                 100 /       747
> Jul 15 13:06:00|00024|coverage|INFO|util_xalloc                412 /      5517
> Jul 15 13:06:00|00025|coverage|INFO|vconn_received             100 /       750
> Jul 15 13:06:00|00026|coverage|INFO|vconn_sent                 100 /       750
> Jul 15 13:06:00|00027|coverage|INFO|hmap_expand                  0 /         1
> Jul 15 13:06:00|00028|coverage|INFO|mac_learning_expired         0 /        13
> Jul 15 13:06:00|00029|coverage|INFO|mac_learning_learned         0 /        36
> Jul 15 13:06:00|00030|coverage|INFO|pstream_open                 0 /         1
> Jul 15 13:06:00|00031|coverage|INFO|vconn_open                   0 /         1
> Jul 15 13:06:00|00032|coverage|INFO|82 events never hit
> Jul 15 13:06:00|00033|timeval|WARN|13 ms poll interval (10 ms user, 10 ms system) is over 11 times the weighted mean interval 1 ms (500 samples)
> Jul 15 13:06:00|00034|timeval|WARN|context switches: 0 voluntary, 1 involuntary
> Jul 15 13:06:00|00035|coverage|INFO|Skipping details of duplicate event coverage for hash=57205ddd in epoch 500
> Jul 15 13:06:01|00036|timeval|WARN|12 ms poll interval (0 ms user, 10 ms system) is over 10 times the weighted mean interval 1 ms (1042 samples)
> Jul 15 13:06:01|00037|coverage|INFO|Skipping details of duplicate event coverage for hash=57205ddd in epoch 1042
> Jul 15 13:06:29|00038|rconn|WARN|tcp:x.x.x.x:50924: connection dropped (Connection reset by peer)
> 
> Can anybody explain what happended?
> The packets are lost, after that my physical switch (which is not an openflow switch) reports huge traffic on all physical ports.
> 
> thx for your answers,
> Lenard
> 
> 2010/7/15 Pásztor Lénárd Zoltán <lenard.pasztor at gmail.com>
> Hi All,
> 
> I'm started testing the openvswitch on a virtualized environment.
> I followed this guide: http://openvswitch.org/cgi-bin/gitweb.cgi?p=openvswitch;a=blob_plain;f=INSTALL.OpenFlow;hb=HEAD
> 
> When I trying to start openflowd I got the following message:
> 
> ovs-openflowd dp0 tcp:x.x.x.x
> Jul 15 11:12:17|00001|openflowd|INFO|Open vSwitch version 1.0.1
> Jul 15 11:12:17|00002|openflowd|INFO|OpenFlow protocol version 0x01
> Jul 15 11:12:17|00003|ofproto|INFO|using datapath ID 0000002320b94533
> Jul 15 11:12:17|00004|rconn|INFO|tcp:x.x.x.x: connecting...
> Jul 15 11:12:17|00005|netdev|WARN|attempted to create a device that may not be created: eth1
> Jul 15 11:12:17|00006|ofproto|WARN|ignoring port eth1 (1) because netdev eth1 cannot be opened (No such device)
> Jul 15 11:12:17|00007|ofproto|WARN|packet-in on unknown port 1
> Jul 15 11:12:17|00008|rconn|INFO|tcp:x.x.x.x: connected
> Jul 15 11:12:17|00009|ofproto|WARN|packet-in on unknown port 1
> 
> more information about dp0:
> 
> ovs-dpctl show
> system at dp0:
>         flows: cur:3, soft-max:512, hard-max:262144
>         ports: cur:2, max:1024
>         groups: max:16
>         lookups: frags:0, hit:61126, missed:21868, lost:6060
>         queues: max-miss:100, max-action:100
>         port 0: dp0 (internal)
>         port 1: eth1
> 
> 
> I have multiple dom0 with multiple physical interfaces (used for avability and bonding) and many hosted guests. I would like to set up virtual switches and networks for our guest OS-es, and manage it centrally if it is possible.
> Will openflow be good for me?
> 
> thx for your answers,
> Lenard
> 
> 
> 
> 
> -- 
> üdv,
> 
>  Lénárd
> _______________________________________________
> discuss mailing list
> discuss at openvswitch.org
> http://openvswitch.org/mailman/listinfo/discuss_openvswitch.org





More information about the discuss mailing list