summaryrefslogtreecommitdiffstats
path: root/filter_plugins
diff options
context:
space:
mode:
authorJoel Diaz <jdiaz@redhat.com>2017-09-12 19:11:37 +0000
committerJoel Diaz <jdiaz@redhat.com>2017-09-12 19:34:19 +0000
commit3ebc0b591f9445b8fde65bda3f76666bad2c8ca3 (patch)
treed707b5da1a13e1ae31d021eab58d534b5211d979 /filter_plugins
parent4acdef4af89bf2ccc43f9643a2e72a969d11ed04 (diff)
downloadopenshift-3ebc0b591f9445b8fde65bda3f76666bad2c8ca3.tar.gz
openshift-3ebc0b591f9445b8fde65bda3f76666bad2c8ca3.tar.bz2
openshift-3ebc0b591f9445b8fde65bda3f76666bad2c8ca3.tar.xz
openshift-3ebc0b591f9445b8fde65bda3f76666bad2c8ca3.zip
move (and rename) get_dns_ip filter into openshift_node_facts
Due to a combination of unexpected behavior when resolving filter plugins in Ansible and recent changes to filter_plugins/openshift_node.py, when trying to run a current version of the openshift_node_facts role on a system with an older version of the openshift-ansible-filter-plugins RPM, the role will fail. File "/usr/share/ansible_plugins/filter_plugins/openshift_node.py", line 30, in get_dns_ip if bool(hostvars['openshift']['common']['use_dnsmasq']): KeyError: 'use_dnsmasq' It's not possible for us to have our current RPM version (presently openshift-ansible-filter-plugins-3.5.120-1.git.0.c60f69a.el7.noarch) and a newer RPM installed and run both current openshift-ansible code and older checked-out/vendored releases (for older clusters/releases). Since only the openshift_node_facts role uses the get_dns_ip filter, move the functionality into a role-specific filter plugin. In addition, rename the filter plugin to 'node_get_dns_ip' because Ansible is prefering the get_dns_ip from the RPM (ie /usr/share/ansible_plugins/filter_plugins/openshift_node.py) over the role-specific filter plugin of the same name. Ansile prefers the filter plugins in /usr/share/ansible_plugins/filter_plugins/* even when /etc/ansible/ansible.cfg is set to something like: filter_plugins = filter_plugins:/usr/share/ansible_plugins/filter_plugins:filter_plugins <--- yes, 'filter_plugins' before and after /usr/share/ansible_plugins/filter_plugins (ansible 2.3.1.0) Renaming the filter plugin function ensures that roles that depend on the old implementation can continue to use it through what the older RPM provides, and the new role can use it's role-specific filter plugin without any variable namespace colisions. Lastly, remove filter_plugins/openshift_node.py since it is now self-contained in roles/openshift_node_facts.
Diffstat (limited to 'filter_plugins')
-rw-r--r--filter_plugins/openshift_node.py32
1 files changed, 0 insertions, 32 deletions
diff --git a/filter_plugins/openshift_node.py b/filter_plugins/openshift_node.py
deleted file mode 100644
index 50c360e97..000000000
--- a/filter_plugins/openshift_node.py
+++ /dev/null
@@ -1,32 +0,0 @@
-#!/usr/bin/python
-# -*- coding: utf-8 -*-
-'''
-Custom filters for use in openshift-node
-'''
-from ansible import errors
-
-
-class FilterModule(object):
- ''' Custom ansible filters for use by openshift_node role'''
-
- @staticmethod
- def get_dns_ip(openshift_dns_ip, hostvars):
- ''' Navigates the complicated logic of when to set dnsIP
-
- In all situations if they've set openshift_dns_ip use that
- For 1.0/3.0 installs we use the openshift_master_cluster_vip, openshift_node_first_master_ip, else None
- For 1.1/3.1 installs we use openshift_master_cluster_vip, else None (product will use kube svc ip)
- For 1.2/3.2+ installs we set to the node's default interface ip
- '''
-
- if not issubclass(type(hostvars), dict):
- raise errors.AnsibleFilterError("|failed expects hostvars is a dict")
-
- # We always use what they've specified if they've specified a value
- if openshift_dns_ip is not None:
- return openshift_dns_ip
- return hostvars['ansible_default_ipv4']['address']
-
- def filters(self):
- ''' returns a mapping of filters to methods '''
- return {'get_dns_ip': self.get_dns_ip}