[ovs-discuss] ICMP_SEQ resets every 70 pings and counts down from a set value

aarggh @aarggh.com aarggh at gmail.com
Mon Aug 25 00:24:59 UTC 2014


Hi all, we have an odd but very specific issue with GRE tunnels we’ve
created using Openvswitch v2.1.2 on Centos 6.5 x64. We have created the
interfaces and GRE tunnels fine as peer the simple and fairly standard
instructions widely available, and once we insert an arp entry, we can ping
endpoints fine, but although everything works just fine at first, we find
that after a few days of uptime that when we ping endpoints, the icmp_seq
goes from low random response times, to a cycle in which every 70 pings it
resets to a set high value, 120ms, and gradually drops down to 22ms over
the next 70 pings.



Further, ping response seems to be padded out where necessary to ensure it
remains a cycle of 70 pings.



Stopping, or restarting Openvswitch doesn’t fix it, but restarting
networking and adding the static arp entry back in does.



We’re at our wits end trying to determine the cause of this strange
behaviour, is there some obscure cache setting in OVS that would cause this?



Regards, Ian
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20140825/e0ab7240/attachment-0002.html>


More information about the discuss mailing list