Packstack: Route to external network disappeared after rebooting controller
On my Newton-based Packstack (all-in-one controller plus two compute nodes), the controller has this route:
$ ip r
...
172.24.4.224/28 dev br-ex proto kernel scope link src 172.24.4.225
All is fine there.
I also deployed an Ocata-based Packstack. After rebooting the controller, I can't access instances through their floating IPs anymore, and indeed, there is no route to 172.24.4.xxx.
I can't confirm that this route existed before, but I see traces in the deployment log file /var/tmp/packstack/latest/openstack-setup.log:
# grep openstack-setup.log
...
ipaddress => 172.24.4.1
ipaddress_br_ex => 172.24.4.1
network_br_ex => 172.24.4.0
2017-03-07 15:40:12::DEBUG::run_setup::626::root:: provision-demo-floatrange: 172.24.4.0/24
My networking skills are not sufficient to figure out where this route disappeared. Any idea where to look?
EDIT: I provisioned a new Packstack controller.
This time, I removed the superfluous eth1 from the controller. As recommended on the quickstart page, firewalld and NW manager disabled, network service enabled. Packstack provisioned like this:
sudo packstack -d --default-password=bla --allinone
Directly after installation, I have no ifcfg-br-ex, but this route exists:
172.24.4.0/24 dev br-ex proto kernel scope link src 172.24.4.1
And:
4: br-ex: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc noqueue state UNKNOWN qlen 1000
link/ether 5a:9d:96:0b:ed:48 brd ff:ff:ff:ff:ff:ff
inet 172.24.4.1/24 scope global br-ex
valid_lft forever preferred_lft forever
After rebooting, no route to 172, and:
9: br-ex: <BROADCAST,MULTICAST> mtu 1500 qdisc noop state DOWN qlen 1000
link/ether 5a:9d:96:0b:ed:48 brd ff:ff:ff:ff:ff:ff
After ip link set br-ex up
and adding the IP address manually, I can access my instance via floating IP. So I do have a good workaround, but somewhere in Packstack the setup of br-ex is not quite watertight.
I submitted a bug description.