Spreading variables across multiple places with a different
naming convention adds confusing about their origin and
default values.
This patch aims to align naming convention as well as defaults
defenition on role level rather then partially in playbook and partially
in role.
Change-Id: I8a665a485cb8028ce7aee62cb9107cd6a482e2e4
At the moment proxy and main mcapi versions are not related in
any way, which means that there can be a conflict between
installed versions of proxy with mcapi.
One of examples was triggered by [1], when proxy was issuing
a requests towards API which does not exist yet in the central
cluster. This resulted in HTTPError: 404 Client Error: Not Found for url
when proxy tried to fetch list of clusters.
[1] 8e375c414a
Change-Id: Iba1f34e7b79e34f23b8169ecd20907832626d540
Makes sense to be consistent and follow suggestion to use group_vars
where applicable.
There is a bug in vexxhost collection which prevents from doing so,
unfortunatelly.
For this to be accurate doc, a patch to vexxhost collection should
land first:
https://github.com/vexxhost/ansible-collection-kubernetes/pull/136
Change-Id: I5f48e913436bf5d6a8d6c0a9f77c58886e451d1f
Right now all roles assume that kibana has been deployed using the
stack and `kibana` group is defined in Ansible.
However if one need just to use journalbeat or filebeat roles to push
data to external Kibana - it's barely possible. Defining `kibana` group
is risky as open doors for installKibana playbook execution.
This patch adds variable kibana_target along with more role-specific
variables that allow to adjust thi behaviour and manually supply
Kibana endpoint to be configured.
Change-Id: Id2a42ae9c6146dcc9e86b15fee36372b95461d20
Enabling ANSIBLE_INJECT_FACT_VARS does result in a performance
regression as each variable loaded to runtime slows down task
execution.
We disable ANSIBLE_INJECT_FACT_VARS in OSA by default for a while,
so ELK role should also be able to run with this setting disabled.
Change-Id: Ibffc09cdb4f9289ddad38211ccb0265642b4321f
In case your hostname contains upper-level symbols, they will be all
lowered in k8s cluster.
With that it's important to ensure that the expected hostname used in
roles is always lowercase one.
Change-Id: I31e483cc7766e26f932984067daee6983122db10
This reverts commit c1ede00a6cb3dfdf776ba4335898ef883474a596.
Reason for revert: This is now included in v1.14.0 of ansible-collection-kubernetes
Change-Id: Icfbe36d45cf39e0b7d8fdf4bbf674711cb34004f
Previously the service would only restart if the systemd unit
was modified. If the venv was changed in-place, this would not
trigger a restart.
Change-Id: I685bedef4d84bf718aaa9064a20d3c1b71daf5bf
Various configuration options for this service are set using
environment variables, including: POD_IP, PROXY_BIND, PROXY_PORT.
This patch exposes a variable which can be used to set these via
the systemd service unit.
Change-Id: I5f649c3894c63e13649f8f12d4dd839b22ba6cfc
In order to respect group_vars install_defaults role moved vars
definition from inlcude_vars to the role defaults. With that now we
need to provide correct path to the defaults during the role import.
Depends-On: https://review.opendev.org/c/openstack/openstack-ansible-plugins/+/923390
Change-Id: I23ad7c955e2d0e267ea6c1aeb50f6be24c515b69
This makes it possible to use the sonobuoy role to test the control
plane cluster as well as workload clusters.
Change-Id: Ia82ebe150774fbd694e74531c3392518887698fd
There is now a release version that includes all the changes necessary
for successful deployment with openstack-ansible.
Change-Id: I5e76bfb1210c5890f6d373acf1a35d0a6d981ff4
Pin prior to OSA Caracal release - ideally this would be a released
version of the collection but OSA requires a recent patch that is
not yet included in a release.
Change-Id: I54acd63939f2e9f87554446d4c33a155a3ffe812
Kubernetes cluster-autoscaler only runs on nodes with a specific label
so add task to set required label(s) on control-plane k8s nodes.
Change-Id: I9bf371ad1ba5d80b7e5a950ac1d60499788d9d0d
