ansible deploy yielding undefined variable

[Jeff Peeler]

[vagrant@operator ansible]$ ansible-playbook -i inventory/multinode -e @/etc/kolla/defaults.yml -e @/etc/kolla/globals.yml -e @/etc/kolla/passwords.yml site.yml

PLAY [haproxy] ****************************************************************

GATHERING FACTS ***************************************************************
ok: [network01]

TASK: [haproxy | Ensuring config directory exists] ****************************
ok: [network01]

TASK: [haproxy | Copying over config(s)] **************************************
fatal: [network01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_eth0'", 'failed': True}
fatal: [network01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_eth0'", 'failed': True}

FATAL: all hosts have already failed -- aborting

PLAY RECAP ********************************************************************
           to retry, use: --limit @/home/vagrant/site.retry

network01 : ok=2 changed=0 unreachable=1 failed=0

----

Another run on a different setup (harm's):

fatal: [support01] => {'msg': "AnsibleUndefinedVariable: One or more undefined variables: 'dict object' has no attribute u'ansible_enp0s8'", 'failed': True}

[Jeff Peeler (jpeeler-z)]

haproxy related

[Jeff Peeler (jpeeler-z)]

I'm closing this based on the fact that I was able to get past this upon changing my network configuration.

[weiyu (weiyu)]

I met the same problem

[weiyu (weiyu)]

        },
        "ansible_eth0": {
            "active": true,
            "device": "eth0",
            "ipv4": {
                "address": "192.168.20.20",
                "netmask": "255.255.255.0",
                "network": "192.168.20.0"
            },
            "ipv6": [
                {
                    "address": "fe80::f816:3eff:fe16:5b17",
                    "prefix": "64",
                    "scope": "link"
                }
            ],
            "macaddress": "fa:16:3e:16:5b:17",
            "module": "virtio_net",
            "mtu": 1500,
            "promisc": false,
            "type": "ether"
        },
        "ansible_eth1": {
            "active": true,
            "device": "eth1",
            "ipv4": {
                "address": "192.168.30.9",
                "netmask": "255.255.255.0",
                "network": "192.168.30.0"
            },
            "ipv6": [
                {
                    "address": "fe80::f816:3eff:fed0:4e8a",
                    "prefix": "64",
                    "scope": "link"

[Jeff Peeler (jpeeler-z)]

So I closed this bug prematurely and also am still seeing this. The network reconfiguration did not stop ansible's complaint.

[Steven Dake (sdake)]

did you set the network interface to a non-eth0 in /etc/kolla/globals.yml as stated in the documentation?

[Steven Dake (sdake)]

This is a confirmed bug that Sam understands how to fix.

For a quick workaround when doing multinode deployment, every node that is in control or compute should also be in the [network] block in the multinode inventory file.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Vladislav Belogrudov (<email address hidden>) on branch: master
Review: https://review.openstack.org/220923
Reason: we need different solution that includes either caching or fixing haproxy role

[Paul Bourke (pauldbourke)]

sdake's workaround did not work for me, Vlad's does https://review.openstack.org/#/c/220923/2 until permanent fix is made.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/220970

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/220970
Committed: https://git.openstack.org/cgit/stackforge/kolla/commit/?id=c68c9d95fca6485c79a607b3716a88e284c7a64e
Submitter: Jenkins
Branch: master

commit c68c9d95fca6485c79a607b3716a88e284c7a64e
Author: Sam Yaple <email address hidden>
Date: Mon Sep 7 12:04:28 2015 +0000

    Gather facts from the hosts before using them

    Haproxy needs to have gathered facts from all hosts that it will use
    information about. In this case it must talk to all of the api hosts
    as well as the database and rabbitmq hosts before generating the
    configuration file.

    Change-Id: I99b7dbebd5a6193e192ee258ddf576d18db90ed7
    Closes-Bug: #1482406