Networking unbelievable slow

asked 2016-03-17 12:06:20 -0500

Hi,

I did a new 3 node (1 controller, 1 network & block, 1 compute) installation, following the instructions from http://docs.openstack.org/liberty/ins... and http://docs.openstack.org/liberty/net... (Classic with Open vSwitch). It looks, like it works, but the virtual network is extremely slow, several minutes for a password prompt via ssh, (except pinging) and therefore unusable. The behaviour is the same if coming from the outside to the floating ip or using "ip netns exec" direct on the network-node. Could somebody be so nice and point me in the right direction what is happening here?

ATB daubler

edit retag flag offensive close merge delete

add a comment

0

answered 2016-03-18 08:24:21 -0500

daubler

43 ●6 ●8 ●13 https://www.eodc.eu/

updated 2016-03-18 08:26:14 -0500

Hi, thanks a lot for this answer. I found out, that the vm did what I tiped in (pinging an otside host) - via tcpdump on the network-node I saw that the packets wer sent out, and came back, I just didn't see anything at the console. So I played around with the tunnel-interfaces (which are bonds) and after setting "ethtool -K tunnelbond tx off" on the compute-node everything worked! Why? I can't say, my network-knowledge doesn't go that far. ATB daubler