Rework the upgrade logic to be more resiliant for multiple upgrades

CONTRIBUTING.md

@@ -6,13 +6,15 @@ Have questions about this document or anything not covered here? Please file a n
 
 ## Table of contents
 
-* [Things to know prior to submitting code](#things-to-know-prior-to-submitting-code)
-* [Submmiting your Work](#submitting-your-work)
-* [Testing](#testing)
-    * [Testing in Docker](#testing-in-docker)
-    * [Testing in Minikube](#testing-in-minikube)
-* [Generating a bundle](#generating-a-bundle)
-* [Reporting Issues](#reporting-issues)
+- [AWX-Operator Contributing Guidelines](#awx-operator-contributing-guidelines)
+  - [Table of contents](#table-of-contents)
+  - [Things to know prior to submitting code](#things-to-know-prior-to-submitting-code)
+  - [Submmiting your work](#submmiting-your-work)
+  - [Testing](#testing)
+      - [Testing in Kind](#testing-in-kind)
+      - [Testing in Minikube](#testing-in-minikube)
+  - [Generating a bundle](#generating-a-bundle)
+  - [Reporting Issues](#reporting-issues)
 
 
 ## Things to know prior to submitting code
@@ -44,12 +46,12 @@ Have questions about this document or anything not covered here? Please file a n
 
 ## Testing
 
-This Operator includes a [Molecule](https://molecule.readthedocs.io/en/stable/)-based test environment, which can be executed standalone in Docker (e.g. in CI or in a single Docker container anywhere), or inside any kind of Kubernetes cluster (e.g. Minikube).
+This Operator includes a [Molecule](https://ansible.readthedocs.io/projects/molecule/)-based test environment, which can be executed standalone in Docker (e.g. in CI or in a single Docker container anywhere), or inside any kind of Kubernetes cluster (e.g. Minikube).
 
 You need to make sure you have Molecule installed before running the following commands. You can install Molecule with:
 
 ```sh
-#> pip install 'molecule[docker]'
+#> python -m pip install molecule-plugins[docker]
 ```
 
 Running `molecule test` sets up a clean environment, builds the operator, runs all configured tests on an example operator instance, then tears down the environment (at least in the case of Docker).

config/manager/kustomization.yaml

@@ -5,9 +5,9 @@ generatorOptions:
   disableNameSuffixHash: true
 
 configMapGenerator:
-- name: awx-manager-config
-  files:
+- files:
   - controller_manager_config.yaml
+  name: awx-manager-config
 
 apiVersion: kustomize.config.k8s.io/v1beta1
 kind: Kustomization

docs/user-guide/database-configuration.md

@@ -56,15 +56,15 @@ If you don't have access to an external PostgreSQL service, the AWX operator can
 
 The following variables are customizable for the managed PostgreSQL service
 
-| Name                                          | Description                                   | Default                            |
-| --------------------------------------------- | --------------------------------------------- | ---------------------------------- |
-| postgres_image                                | Path of the image to pull                     | postgres:15                        |
-| postgres_init_container_resource_requirements | Database init container resource requirements | requests: {cpu: 10m, memory: 64Mi} |
-| postgres_resource_requirements                | PostgreSQL container resource requirements    | requests: {cpu: 10m, memory: 64Mi} |
-| postgres_storage_requirements                 | PostgreSQL container storage requirements     | requests: {storage: 8Gi}           |
-| postgres_storage_class                        | PostgreSQL PV storage class                   | Empty string                       |
-| postgres_data_path                            | PostgreSQL data path                          | `/var/lib/postgresql/data/pgdata`  |
-| postgres_priority_class                       | Priority class used for PostgreSQL pod        | Empty string                       |
+| Name                                          | Description                                   | Default                                 |
+| --------------------------------------------- | --------------------------------------------- | --------------------------------------- |
+| postgres_image                                | Path of the image to pull                     | quay.io/sclorg/postgresql-15-c9s:latest |
+| postgres_init_container_resource_requirements | Database init container resource requirements | requests: {cpu: 10m, memory: 64Mi}      |
+| postgres_resource_requirements                | PostgreSQL container resource requirements    | requests: {cpu: 10m, memory: 64Mi}      |
+| postgres_storage_requirements                 | PostgreSQL container storage requirements     | requests: {storage: 8Gi}                |
+| postgres_storage_class                        | PostgreSQL PV storage class                   | Empty string                            |
+| postgres_data_path                            | PostgreSQL data path                          | `/var/lib/postgresql/data/pgdata`       |
+| postgres_priority_class                       | Priority class used for PostgreSQL pod        | Empty string                            |
 
 Example of customization could be:
 

molecule/default/tasks/awx_replicas_test.yml

@@ -61,4 +61,4 @@
       expected_web_replicas: 3
       expected_task_replicas: 3
   tags:
-    - replicas
+  - replicas

roles/installer/tasks/database_configuration.yml

@@ -106,14 +106,38 @@
   set_fact:
     managed_database: "{{ pg_config['resources'][0]['data']['type'] | default('') | b64decode == 'managed' }}"
 
-- name: Get the old postgres pod information
+# It is possible that N-2 postgres pods may still be present in the namespace from previous upgrades.
+# So we have to take that into account and preferentially set the most recent one.
+- name: Get the old postgres pod (N-1)
   k8s_info:
     kind: Pod
     namespace: "{{ ansible_operator_meta.namespace }}"
-    name: "{{ ansible_operator_meta.name }}-postgres-0"
     field_selectors:
       - status.phase=Running
-  register: old_postgres_pod
+  register: _running_pods
+
+- block:
+  - name: Filter pods by name
+    set_fact:
+      filtered_old_postgres_pods: "{{ _running_pods.resources |
+        selectattr('metadata.name', 'match', ansible_operator_meta.name + '-postgres.*-0') |
+        rejectattr('metadata.name', 'search', '-' + supported_pg_version | string + '-0') |
+        list }}"
+
+  # Sort pods by name in reverse order (most recent PG version first) and set
+  - name: Set info for previous postgres pod
+    set_fact:
+      sorted_old_postgres_pods: "{{ filtered_old_postgres_pods |
+        sort(attribute='metadata.name') |
+        reverse }}"
+    when: filtered_old_postgres_pods | length
+
+
+  - name: Set info for previous postgres pod
+    set_fact:
+      old_postgres_pod: "{{ sorted_old_postgres_pods | first }}"
+    when: filtered_old_postgres_pods | length
+  when: _running_pods.resources | length
 
 - name: Look up details for this deployment
   k8s_info:
@@ -123,6 +147,13 @@
     namespace: "{{ ansible_operator_meta.namespace }}"
   register: this_awx
 
+# If this deployment has been upgraded before or if upgrade has already been started, set this var
+- name: Set previous PG version var
+  set_fact:
+    _previous_upgraded_pg_version: "{{ this_awx['resources'][0]['status']['upgradedPostgresVersion'] | default(false) }}"
+  when:
+    - "'upgradedPostgresVersion' in this_awx['resources'][0]['status']"
+
 - name: Check if postgres pod is running an older version
   block:
     - name: Set path to PG_VERSION file for given container image
@@ -132,22 +163,24 @@
     - name: Get old PostgreSQL version
       k8s_exec:
         namespace: "{{ ansible_operator_meta.namespace }}"
-        pod: "{{ ansible_operator_meta.name }}-postgres-0"
+        pod: "{{ old_postgres_pod['metadata']['name'] }}"
         command: |
           bash -c """
           cat {{ path_to_pg_version }}
           """
       register: _old_pg_version
 
+    - debug:
+        msg: "--- Upgrading from {{ old_postgres_pod['metadata']['name'] | default('NONE')}} Pod ---"
+
     - name: Upgrade data dir from old Postgres to {{ supported_pg_version }} if applicable
       include_tasks: upgrade_postgres.yml
       when:
-        - _old_pg_version.stdout | default('0') | trim < supported_pg_version
+        - (_old_pg_version.stdout | default(0) | int ) < supported_pg_version
   when:
     - managed_database
-    - "'upgradedPostgresVersion' is in this_awx['resources'][0]['status']"
-    - (this_awx['resources'][0]['status']['upgradedPostgresVersion'] | default(false)) | ternary (this_awx['resources'][0]['status']['upgradedPostgresVersion'] < supported_pg_version, false)
-    - old_postgres_pod['resources'] | length  # upgrade is complete and old pg pod has been removed
+    - (_previous_upgraded_pg_version | default(false)) | ternary(_previous_upgraded_pg_version < supported_pg_version, true)
+    - old_postgres_pod | length  # If empty, then old pg pod has been removed and we can assume the upgrade is complete
 
 - block:
     - name: Create Database if no database is specified

roles/installer/tasks/update_status.yml

@@ -111,5 +111,5 @@
     name: "{{ ansible_operator_meta.name }}"
     namespace: "{{ ansible_operator_meta.namespace }}"
     status:
-      upgradedPostgresVersion: "{{ upgraded_postgres_version }}"
+      upgradedPostgresVersion: "{{ upgraded_postgres_version | string }}"
   when: upgraded_postgres_version is defined

roles/installer/tasks/upgrade_postgres.yml

@@ -62,9 +62,19 @@
   set_fact:
     postgres_pod_name: "{{ postgres_pod['resources'][0]['metadata']['name'] }}"
 
+- name: Get the name of the service for the old postgres pod
+  k8s_info:
+    kind: Service
+    namespace: "{{ ansible_operator_meta.namespace }}"
+    label_selectors:
+      - "app.kubernetes.io/component=database"
+      - "app.kubernetes.io/instance={{ old_postgres_pod.metadata.labels['app.kubernetes.io/instance'] }}"
+      - "app.kubernetes.io/managed-by=awx-operator"
+  register: old_postgres_svc
+
 - name: Set full resolvable host name for postgres pod
   set_fact:
-    resolvable_db_host: "{{ ansible_operator_meta.name }}-postgres.{{ ansible_operator_meta.namespace }}.svc"  # yamllint disable-line rule:line-length
+    resolvable_db_host: "{{ old_postgres_svc['resources'][0]['metadata']['name'] }}.{{ ansible_operator_meta.namespace }}.svc"  # yamllint disable-line rule:line-length
   no_log: "{{ no_log }}"
 
 - name: Set pg_dump command
@@ -110,23 +120,32 @@
     kind: StatefulSet
     api_version: v1
     namespace: "{{ ansible_operator_meta.namespace }}"
-    name: "{{ ansible_operator_meta.name }}-postgres"
+    name: "{{ item }}"
     state: absent
     wait: true
+  loop:
+    - "{{ ansible_operator_meta.name }}-postgres"
+    - "{{ ansible_operator_meta.name }}-postgres-13"
 
 - name: Remove old Postgres Service
   k8s:
     kind: Service
     api_version: v1
     namespace: "{{ ansible_operator_meta.namespace }}"
-    name: "{{ ansible_operator_meta.name }}-postgres"
+    name: "{{ item }}"
     state: absent
+  loop:
+    - "{{ ansible_operator_meta.name }}-postgres"
+    - "{{ ansible_operator_meta.name }}-postgres-13"
 
 - name: Remove old persistent volume claim
   k8s:
     kind: PersistentVolumeClaim
     api_version: v1
     namespace: "{{ ansible_operator_meta.namespace }}"
-    name: "postgres-{{ ansible_operator_meta.name }}-postgres-0"
+    name: "{{ item }}"
     state: absent
+  loop:
+    - "postgres-{{ ansible_operator_meta.name }}-postgres-0"
+    - "postgres-{{ ansible_operator_meta.name }}-postgres-13-0"
   when: postgres_keep_pvc_after_upgrade

roles/installer/vars/main.yml

@@ -5,3 +5,5 @@ ldap_cacert_ca_crt: ''
 bundle_ca_crt: ''
 projects_existing_claim: ''
 supported_pg_version: 15
+_previous_upgraded_pg_version: 0
+old_postgres_pod: []