| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
on inventory/playbook variables for openshift_hostname
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- Remove default value for openshift_hostname and make it required
- Remove workarounds that are no longer needed
- Remove resources parameter from openshift_register_node module
- pre-create node certificates for each node before registering node
- distribute created node certificates to each node
- Move node registration logic to a new openshift_register_nodes role
- This is because we now have to run the steps on a master as opposed to on
the nodes like we were previously doing.
- Rename openshift_register_node module to kubernetes_register_node, one more
step to genericizing enough for upstreaming, however there are still plenty
of openshift specific commands that still need to be genericized.
|
| |
|
|\
| |
| | |
Random cleanup
|
| | |
|
| | |
|
|\ \
| | |
| | | |
Conditionally set --nodes on master
|
| | |
| | |
| | |
| | |
| | | |
- only add --nodes option to /etc/sysconfig/openshift-master when
openshift_node_ips is not an empty list.
|
|/ /
| |
| |
| | |
- missing leading 0 on mode
|
| |
| |
| |
| |
| | |
- Fix missed references to old firewall scripts
- Fix variable name references that didn't get updated
|
|\ \
| | |
| | | |
Do not set KUBECONFIG for root user
|
| |/
| |
| |
| |
| |
| |
| | |
- instead of setting KUBECONFIG, copy the admin kubeconfig to
/root/.kube/.kubeconfig in the openshift_master and openshift_node roles
- pause for 30 seconds if the openshift-master service has changed state,
since the file we are copying is generated by the master
|
|/
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
- openshift_node_ips now defaults to []
- Previously an empty --nodes in /etc/sysconfig/master would result in the
master creating a node for the localhost. The latest Origin and OSE builds
now only create the implicit localhost node if run as openshift, not
openshift-master. We can now safely default to setting no nodes in
/etc/sysconfig/master and having nodes register themselves with the master
when they come up via the 'Register node (if not already registered)' task
in roles/openshift_node/tasks/main.yml)
- This had an associated change for the byo scripts that had not been merged
into master yet, but this PR changes the behavior of the openshift_master
role to not fail if openshift_node_ips is not set. This also prevents having
the openshift_master service restarted when a node is added.
|
| |
|
|\
| |
| | |
Prefer YAML style datastructures over JSON
|
| |
| |
| |
| | |
- Switch JSON style datastructures to YAML for debuggability
|
| |
| |
| |
| |
| | |
- for use anywhere the hostname is used that would be dependent on
the openshift_hostname_workaround setting.
|
|/
|
|
|
| |
- use openshift_bind_ip for hostname when openshift_hostname_workaround is true
- defaults to true to maintain current behavior.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
- move common openshift logic into openshift_common
- set openshift_common as a dependency for openshift_node and openshift_master
- rename role variables to openshift_* to be more descriptive
- start recording local_facts on the openshift hosts
- clean up firewalld config to be a bit more dry
- Update firewall ports for https, make sure http rules are removed
- Replace references to ansible_eth0.ipv4.address with
ansible_default_ipv4.address
|
| |
|
|
|
|
| |
multi-line so we get better errors from ansible.
|
|
|
|
|
|
|
|
|
|
|
|
| |
- roles/base_os: Without this, the root user would need to manually configure
this variable before attempting to run any osc commands
- roles/base_os: Cleanup the firewall service definition and only pause when
the service
state changes.
- roles/openshift_master: use Akram's suggestion of simplifying the firewall
config
- roles/openshift_master: explicitly disable previously exposed ports that are
no longer exposed (8080/tcp I'm looking at you).
|
| |
|
|
|
|
| |
- Update playbooks to support latest code
|
|
|
|
| |
yaml dictionaries as demonstrated by the ansible documentation.
|
|
|
|
|
| |
* openshift forces resolving all minions to hostnames to query etcd keys
rather than using the IP address
|
| |
|
| |
|
|
|