lbaas haproxy "backend has no server available" spamming terminal via systemd

[Drew Freiberger]

We are finding the haproxy for lbaas agents that are in a state of not having instances related to them on the backend are spamming user ssh terminals.

This should be sufficient to be in haproxy logs. It makes it very difficult to work on the units.

Settings in n-gw charm are

jujumanage@cmg01z00infr001:~$ juju config neutron-gateway debug
true
jujumanage@cmg01z00infr001:~$ juju config neutron-gateway verbose
false

I wonder if debug is sending to STDERR as well as the log. Even so, I'd rather systemd be configured to not forward that on to terminals.

Please see http://pastebin.ubuntu.com/26180641/ for details of related processes running and messages from systemd-journald.

Running on xenial
python-neutron-lbaas 2:10.0.1-0ubuntu1~cloud0
haproxy 1.6.3-1ubuntu0.1

17.08 charms