etcd fails when kuryr is disabled

[Mark Goddard]

Steps to reproduce:

Enable etcd and not kuryr, then deploy:

enable_etcd: yes
enable_kuryr: no

kolla-ansible deploy

Expected results:

Successful deployment, working etcd.

Actual results:

Successful deployment, but etcd does not work.

This is because the environment variable ETCD_INITIAL_CLUSTER contains compute nodes, which are not running etcd.

Compute nodes are in the etcd group to allow for kuryr, which needs etcd to be running on compute nodes (why???). When kuryr is disabled, etcd is not deployed on these hosts (using host_in_groups). However, the ETCD_INITIAL_CLUSTER environment variable always includes all hosts in the etcd group.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/586479
Committed: https://git.openstack.org/cgit/openstack/kolla-ansible/commit/?id=b9208cd178cab27438831b0d16b337c54c53090b
Submitter: Zuul
Branch: master

commit b9208cd178cab27438831b0d16b337c54c53090b
Author: Mark Goddard <email address hidden>
Date: Thu Jul 26 18:58:25 2018 +0100

    Fix etcd when Kuryr is disabled

    This is because the environment variable ETCD_INITIAL_CLUSTER contains
    compute nodes, which are not running etcd.

    Compute nodes are in the etcd group to allow for kuryr, which needs etcd
    to be running on compute nodes (why???). When kuryr is disabled, etcd is
    not deployed on these hosts (using host_in_groups). However, the
    ETCD_INITIAL_CLUSTER environment variable always includes all hosts in
    the etcd group.

    This change sets ETCD_INITIAL_CLUSTER correctly in both cases.

    Change-Id: I11f812d3fdacf0eeb8236f9a39d673aaa480dd26
    Closes-Bug: #1783963

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/kolla-ansible 7.0.0.0rc1 release candidate.