libvirt provider hostname resolution not working

[richm]

Host: Fedora 23
using vagrant with --provider=libvirt with

  "libvirt": {
    "box_name": "centos7_inst",
    "box_url": "http://mirror.openshift.com/pub/vagrant/boxes/openshift3/centos7_libvirt_inst.box"
  },

NOTE: Everything works fine with the aws provider, this is only an issue with libvirt/virtualbox.

After successfully running vagrant sync-origin, trying to run test-origin-aggregated-logging:

$ vagrant test-origin-aggregated-logging -d  --env OS_ROOT=/data/src/github.com/openshift/origin --env DEBUG_FAILURES=true

This will run the logging.sh script: https://github.com/openshift/origin-aggregated-logging/blob/master/hack/testing/logging.sh#L69 and

os::util::environment::setup_all_server_vars "origin-aggregated-logging/"
os::util::environment::use_sudo
reset_tmp_dir

os::log::start_system_logger

export KUBELET_HOST=$(hostname)

configure_os_server
start_os_server

start_os_server fails here:

wait_for_url "${API_SCHEME}://${API_HOST}:${API_PORT}/healthz" "apiserver: " 0.25 80

API_HOST is openshiftdev.local API_PORT is 10250

$ getent ahostsv4 openshiftdev.local
127.0.0.1       STREAM openshiftdev.local
127.0.0.1       DGRAM  
127.0.0.1       RAW    

$ grep 127.0.0.1 /etc/hosts
127.0.0.1   openshiftdev.local openshiftdev localhost localhost.localdomain localhost4 localhost4.localdomain4

$ netstat -nlt
Active Internet connections (only servers)
Proto Recv-Q Send-Q Local Address           Foreign Address         State      
tcp        0      0 192.168.121.216:4001    0.0.0.0:*               LISTEN     
tcp        0      0 192.168.121.216:10250   0.0.0.0:*               LISTEN     
tcp        0      0 127.0.0.1:6060          0.0.0.0:*               LISTEN     
tcp        0      0 127.0.0.1:53            0.0.0.0:*               LISTEN     
tcp        0      0 192.168.121.216:8053    0.0.0.0:*               LISTEN     
tcp        0      0 0.0.0.0:22              0.0.0.0:*               LISTEN     
tcp        0      0 192.168.121.216:7001    0.0.0.0:*               LISTEN     
tcp        0      0 127.0.0.1:25            0.0.0.0:*               LISTEN     
tcp        0      0 192.168.121.216:8443    0.0.0.0:*               LISTEN     
tcp6       0      0 :::22                   :::*                    LISTEN     
tcp6       0      0 ::1:25                  :::*                    LISTEN     

$ getent hosts 192.168.121.216
192.168.121.216 openshiftdev.local
172.17.0.1      openshiftdev.local

The wait_for_url curl is going to fail because it is trying to talk to the server running on localhost, but the server is only listening to the 192.168.121.216 address. I think the server should listen to all interfaces - but when I set --env ALL_IP_ADDRESSES=0.0.0.0 the script fails with

/data/src/github.com/openshift/origin/hack/util.sh: line 51: SERVER_HOSTNAME_LIST: unbound variable

Perhaps this isn't a bug - perhaps there is some documentation about how to set ALL_IP_ADDRESSES and SERVER_HOSTNAME_LIST to make this work properly?

[richm]

vagrant-aws works - here is some information from an aws machine:

# openshift start --print-ip
172.18.1.188
# ip a - eth0 is 172.18.1.188
# hostname
ip-172-18-1-188
# hostname -f
ip-172-18-1-188.ec2.internal
# getent ahostsv4 ip-172-18-1-188
172.18.1.188
# getent hosts 172.18.1.188
ip-172-18-1-188.ec2.internal
# there is nothing added to /etc/hosts
# cat /etc/resolv.conf
search ec2.internal
nameserver 172.18.0.2

openshift is listening only to 172.18.1.188, but in this case, fortunately, all of the reverse and forward IP/hostname resolution is working correctly.