Comment 37 for bug 1794991

Revision history for this message
Alexandre Perreault (alexperreault) wrote :

Here is an example of what I described in my previous comment. Our biggest problem is with table 22 as it causes outages for our clients.

Here is an example of flows for a vxlan network on a host. As you can see there is no table 22 at all.

docker exec -ti openvswitch_vswitchd ovs-ofctl dump-flows br-tun | grep 0x2b4
 cookie=0xfdd60859f2658796, duration=2037831.278s, table=4, n_packets=245, n_bytes=52188, priority=1,tun_id=0x2b4 actions=mod_vlan_vid:181,resubmit(,9)
 cookie=0xfdd60859f2658796, duration=304285.121s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=181,dl_dst=fa:16:3e:6d:55:e1 actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490a"
 cookie=0xfdd60859f2658796, duration=299805.493s, table=20, n_packets=34, n_bytes=6528, priority=2,dl_vlan=181,dl_dst=fa:16:3e:99:cc:4c actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490c"
 cookie=0xfdd60859f2658796, duration=293855.521s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=181,dl_dst=fa:16:3e:f2:2f:7d actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490a"
 cookie=0xfdd60859f2658796, duration=293234.777s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=181,dl_dst=fa:16:3e:38:08:be actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490e"

Then on another host for the SAME vxlan network we have the opposite. There is a table 22 but all the tunnels are incorrect (They dont match the table 20 flows/tunnels).

docker exec -ti openvswitch_vswitchd ovs-ofctl dump-flows br-tun | grep 0x2b4
 cookie=0xe4add62aa34905c9, duration=2040773.745s, table=4, n_packets=0, n_bytes=0, priority=1,tun_id=0x2b4 actions=mod_vlan_vid:669,resubmit(,9)
 cookie=0xe4add62aa34905c9, duration=318238.768s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=669,dl_dst=fa:16:3e:af:9d:f4 actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490d"
 cookie=0xe4add62aa34905c9, duration=318227.433s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=669,dl_dst=fa:16:3e:4c:c0:93 actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490d"
 cookie=0xe4add62aa34905c9, duration=304524.808s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=669,dl_dst=fa:16:3e:6d:55:e1 actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490a"
 cookie=0xe4add62aa34905c9, duration=300045.175s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=669,dl_dst=fa:16:3e:99:cc:4c actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490c"
 cookie=0xe4add62aa34905c9, duration=294095.211s, table=20, n_packets=0, n_bytes=0, priority=2,dl_vlan=669,dl_dst=fa:16:3e:f2:2f:7d actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490a"
 cookie=0xe4add62aa34905c9, duration=2040773.792s, table=22, n_packets=8, n_bytes=720, priority=1,dl_vlan=669 actions=strip_vlan,load:0x2b4->NXM_NX_TUN_ID[],output:"vxlan-0aa2490f",output:"vxlan-0aa24911",output:"vxlan-0aa24910",output:"vxlan-0aa2490b",output:"vxlan-0aa2490d"

These are outputs from today from the same environment I talked about on december 19th...

I can also provide similar outputs for other networks on different environments. In another environment, on one host, table 22 only has one tunnel when it should have four, and on another host but for the same network, table 22 contains 12 tunnels! Again there should only be four...

I dont understand why these inconsistencies happen.