[Openstack] Floating IP issues in multiple cloud installs.

Brian Haley haleyb.dev at gmail.com
Tue Jun 20 01:29:33 UTC 2017


On 06/19/2017 08:51 AM, Ken D'Ambrosio wrote:
> Hi, all.  We've got two Canonical Newton installs using VLANs and we're 
> having intermittent issues we simply can't figure out.  (Note that a 
> third installation using flat networks is not having this issue.) 
> Floating IPs set up and work... sporadically.
> 
> * Stateful connections (e.g., SSH) often drop after seconds of use to 
> both the FIP and when SSH'd in from the
> * We see RSTs in our TCP dumps
> * Pings work for a while, then don't.
> * We see lots of ARP requests -- even one right after another -- to 
> resolve hosts on the internal subnets:
> 05:43:25.859448 ARP, Request who-has 80.0.0.3 tell 80.0.0.1, length 28
> 05:43:25.859563 ARP, Reply 80.0.0.3 is-at fa:16:3e:28:af:77, length 28
> 05:43:25.964417 ARP, Request who-has 80.0.0.3 tell 80.0.0.1, length 46
> 05:43:25.964572 ARP, Reply 80.0.0.3 is-at fa:16:3e:28:af:77, length 28
> 05:43:26.963989 ARP, Request who-has 80.0.0.3 tell 80.0.0.1, length 46
> 05:43:26.964156 ARP, Reply 80.0.0.3 is-at fa:16:3e:28:af:77, length 28

Was that run with '-i any' or on a single interface?  I would check the 
ARP cache to make sure things entries are in a complete/reachable state. 
  Or even syslog for any other errors.

> 80.0.0.1 is the qrouter.  I can't imagine why it asked -- and was ACK'd 
> in each case -- three times in just over a second.  In hindsight, I 
> should have checked to have seen if the ACK showed up in the qrouter's 
> ARP table.  Next time...

I'd also specify -e to tcpdump to see the MACs involved.  Possibly there 
is something else configured with the same IP on the VLAN (shouldn't 
happen, but worth checking).

-Brian



More information about the Openstack mailing list