Add prepare_client_pod stage for warm shiftstackclient bootstrap - #43
Conversation
| vars: | ||
| _ic: "{{ prepare_client_pod_install_config_slurp.content | b64decode | from_yaml }}" | ||
| ansible.builtin.set_fact: | ||
| api_accessible_ip: "{{ _ic.platform.openstack.apiFloatingIP }}" |
There was a problem hiding this comment.
This unconditionally reads platform.openstack.apiFloatingIP and ingressFloatingIP but the proxy install-config use apiVIP and ingressVIP instead.
As a result, enabling prepare_clint_pod for the existing *ipi-proxy jobs will fail here with an undefined-variable error. Please handle proxy installation using the same address election logic as the existing prepare role or explixitly skil/fail this stage with clear reason for unsupported proxy deployments.
There was a problem hiding this comment.
Good catch — fixed. We now prefer apiFloatingIP / ingressFloatingIP (standard IPI), then fall back to apiVIP/ingressVIP (or apiVIPs/ingressVIPs, taking the first entry when a list). If neither form resolves, the stage fails with an explicit assert instead of an undefined-variable error.
| failed_when: false | ||
|
|
||
| - name: Bootstrap oc and kubectl from the OpenShift mirror | ||
| when: prepare_client_pod_oc_which.rc != 0 |
There was a problem hiding this comment.
This condition checks only oc exists. So the bootstrap is skipped when oc exists, but kutectl is absent. Please check both commands before skipping the bootstrap block.
There was a problem hiding this comment.
Agreed — fixed. Bootstrap now runs when oc or kubectl is missing (which checks for both), so a partial install no longer skips installing the missing binary.
4dd0fbd to
604e6bd
Compare
| tasks_from: create_installer_group.yml | ||
| when: "not deploy_installer_host" | ||
|
|
||
| - name: Prepare the shiftstackclient pod when full prepare was skipped |
There was a problem hiding this comment.
Could we run prepare_client_pod before create_installer_group.yml? On a recreated client pod, the OpenStack credentials have not been restored when create_installer_group.yml calls openstack.cloud.server_info. That task suppresses the error and treats an existing installer VM as absent, adding localhost to the installer inventory. Restoring the credentials afterward cannot correct that inventory.
There was a problem hiding this comment.
Good catch. Moved the prepare_client_pod import ahead of the play that runs create_installer_group.yml, so OpenStack credentials are restored before server_info looks up the installer VM on a recreated client pod.
| remote_src: true | ||
| mode: preserve | ||
|
|
||
| - name: Fall back to original-config OpenStack clouds when PVC copy is absent |
There was a problem hiding this comment.
Could we verify that the fallback clouds.yaml contains clouds[user_cloud] before accepting it? The prepare role adds that cloud to .config/openstack/clouds.yaml and copies it to osp_config_dir; it does not add it to .original-config. If the PVC copy is missing and the original config contains only the admin cloud, this branch succeeds without restoring the shiftstack cloud, leaving the failure for later OpenStack tasks.
There was a problem hiding this comment.
Agreed. Before accepting the original-config fallback we now load that clouds.yaml and assert clouds[user_cloud] is present. After either PVC or fallback restore we also assert the same key on ~/.config/openstack/clouds.yaml, so an admin-only original config cannot silently succeed.
604e6bd to
0d2f243
Compare
When cifmw recreates shiftstackclient and jobs omit prepare (Pipeline B run-tests), the empty pod lacks oc, clouds.shiftstack, /etc/hosts, and resources.yml. Add a lightweight stage that restores only that subset from the PVC / install-config without re-running full prepare. Address review feedback: - Assert a clouds.yaml source exists before creating ~/.config/openstack - Bootstrap when oc or kubectl is missing - Resolve API/apps from FloatingIP with VIP/VIPs fallback for proxy IC - Run prepare_client_pod before create_installer_group - Reject original-config fallback unless clouds[user_cloud] is present - Add YAML document start for ansible-lint Co-authored-by: Cursor <cursoragent@cursor.com>
0d2f243 to
0f31965
Compare
|
/lgtm |
|
/approve |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: ekuris-redhat The full list of commands accepted by this bot can be found here. The pull request process is described here DetailsNeeds approval from an approver in each of these files:
Approvers can indicate their approval by writing |
3 participants
Summary
prepare_client_podstage for warm testing when fullprepareis skipped.oc/kubectlfrom the OpenShift mirror~/.config/openstackfrom the PVCosp_config_dir(fallback: original-config)/etc/hostsAPI/apps entries frominstall-config.yamlFIPsartifacts/resources.ymlwhen missingMotivation
Callers that omit
prepare(to avoid re-initializing a warm guest) still need a way to rehydrate the empty client pod. Keeping that subset in shiftstack-qa avoids duplicating prepare knowledge in cifmw.Usage
Include
prepare_client_podin thestageslist (or stages override) before verification/test stages.Test plan
stagesoc,clouds.shiftstack, guest/etc/hosts, andresources.ymlare available before verificationpreparepath remains unchangedMade with Cursor