Merge pull request #2052 from sdodson/metrics

Metrics improvements

4 files changed, 228 insertions, 58 deletions

diff --git a/roles/openshift_metrics/handlers/main.yml b/roles/openshift_metrics/handlers/main.yml
new file mode 100644
index 000000000..edb7369de
--- /dev/null
+++ b/roles/openshift_metrics/handlers/main.yml
@@ -0,0 +1,31 @@
+---
+- name: restart master
+  service: name={{ openshift.common.service_type }}-master state=restarted
+  when: (openshift.master.ha is not defined or not openshift.master.ha | bool) and (not (master_service_status_changed | default(false) | bool))
+  notify: Verify API Server
+
+- name: restart master api
+  service: name={{ openshift.common.service_type }}-master-api state=restarted
+  when: (openshift.master.ha is defined and openshift.master.ha | bool) and (not (master_api_service_status_changed | default(false) | bool)) and openshift.master.cluster_method == 'native'
+  notify: Verify API Server
+
+- name: restart master controllers
+  service: name={{ openshift.common.service_type }}-master-controllers state=restarted
+  when: (openshift.master.ha is defined and openshift.master.ha | bool) and (not (master_controllers_service_status_changed | default(false) | bool)) and openshift.master.cluster_method == 'native'
+
+- name: Verify API Server
+  # Using curl here since the uri module requires python-httplib2 and
+  # wait_for port doesn't provide health information.
+  command: >
+    curl --silent
+    {% if openshift.common.version_gte_3_2_or_1_2 | bool %}
+    --cacert {{ openshift.common.config_base }}/master/ca-bundle.crt
+    {% else %}
+    --cacert {{ openshift.common.config_base }}/master/ca.crt
+    {% endif %}
+    {{ openshift.master.api_url }}/healthz/ready
+  register: api_available_output
+  until: api_available_output.stdout == 'ok'
+  retries: 120
+  delay: 1
+  changed_when: false
diff --git a/roles/openshift_metrics/tasks/install.yml b/roles/openshift_metrics/tasks/install.yml
new file mode 100644
index 000000000..2fbb7d606
--- /dev/null
+++ b/roles/openshift_metrics/tasks/install.yml
@@ -0,0 +1,114 @@
+---
+
+- name: Test if metrics-deployer service account exists
+  command: >
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace=openshift-infra
+    get serviceaccount metrics-deployer -o json
+  register: serviceaccount
+  changed_when: false
+  failed_when: false
+
+- name: Create metrics-deployer Service Account
+  shell: >
+    echo {{ metrics_deployer_sa | to_json | quote }} |
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    create -f -
+  when: serviceaccount.rc == 1
+
+- name: Test edit permissions
+  command: >
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    get rolebindings -o jsonpath='{.items[?(@.metadata.name == "edit")].userNames}'
+  register: edit_rolebindings
+  changed_when: false
+
+- name: Add edit permission to the openshift-infra project to metrics-deployer SA
+  command: >
+    {{ openshift.common.admin_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    policy add-role-to-user edit
+    system:serviceaccount:openshift-infra:metrics-deployer
+  when: "'system:serviceaccount:openshift-infra:metrics-deployer' not in edit_rolebindings.stdout"
+
+- name: Test cluster-reader permissions
+  command: >
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    get clusterrolebindings -o jsonpath='{.items[?(@.metadata.name == "cluster-reader")].userNames}'
+  register: cluster_reader_clusterrolebindings
+  changed_when: false
+
+- name: Add cluster-reader permission to the openshift-infra project to heapster SA
+  command: >
+    {{ openshift.common.admin_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    policy add-cluster-role-to-user cluster-reader
+    system:serviceaccount:openshift-infra:heapster
+  when: "'system:serviceaccount:openshift-infra:heapster' not in cluster_reader_clusterrolebindings.stdout"
+
+- name: Create metrics-deployer secret
+  command: >
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    secrets new metrics-deployer nothing=/dev/null
+  register: metrics_deployer_secret
+  changed_when: metrics_deployer_secret.rc == 0
+  failed_when: "metrics_deployer_secret.rc == 1 and 'already exists' not in metrics_deployer_secret.stderr"
+
+# TODO: extend this to allow user passed in certs or generating cert with
+# OpenShift CA
+- name: Build metrics deployer command
+  set_fact:
+    deployer_cmd: "{{ openshift.common.client_binary }} process -f \
+      {{ metrics_template_dir }}/metrics-deployer.yaml -v \
+      HAWKULAR_METRICS_HOSTNAME={{ metrics_hostname }},USE_PERSISTENT_STORAGE={{metrics_persistence | string | lower }},METRIC_DURATION={{ openshift.hosted.metrics.duration }},METRIC_RESOLUTION={{ openshift.hosted.metrics.resolution }}{{ image_prefix }}{{ image_version }},MODE={{ deployment_mode }} \
+        | {{ openshift.common.client_binary }} --namespace openshift-infra \
+        --config={{ openshift_metrics_kubeconfig }} \
+        create -f -"
+
+- name: Deploy Metrics
+  shell: "{{ deployer_cmd }}"
+  register: deploy_metrics
+  failed_when: "'already exists' not in deploy_metrics.stderr and deploy_metrics.rc != 0"
+  changed_when: deploy_metrics.rc == 0
+
+- set_fact:
+    deployer_pod: "{{ deploy_metrics.stdout[1:2] }}"
+
+# TODO: re-enable this once the metrics deployer validation issue is fixed
+# when using dynamically provisioned volumes
+- name: "Wait for image pull and deployer pod"
+  shell: >
+    {{ openshift.common.client_binary }}
+    --namespace openshift-infra
+    --config={{ openshift_metrics_kubeconfig }}
+    get {{ deploy_metrics.stdout }}
+  register: deploy_result
+  until: "{{ 'Completed' in deploy_result.stdout }}"
+  failed_when: "{{ 'Completed' not in deploy_result.stdout }}"
+  retries: 60
+  delay: 10
+
+- name: Configure master for metrics
+  modify_yaml:
+    dest: "{{ openshift.common.config_base }}/master/master-config.yaml"
+    yaml_key: assetConfig.metricsPublicURL
+    yaml_value: "https://{{ metrics_hostname }}/hawkular/metrics"
+  notify: restart master
+
+- name: Store metrics public_url
+  openshift_facts:
+    role: master
+    local_facts:
+      metrics_public_url: "https://{{ metrics_hostname }}/hawkular/metrics"
+  when: deploy_result | changed
diff --git a/roles/openshift_metrics/tasks/main.yaml b/roles/openshift_metrics/tasks/main.yaml
index 43b85204a..8a6712468 100644
--- a/roles/openshift_metrics/tasks/main.yaml
+++ b/roles/openshift_metrics/tasks/main.yaml
@@ -1,64 +1,88 @@
 ---
-- name: Copy Configuration to temporary conf
-  command: >
-    cp {{ openshift.common.config_base }}/master/admin.kubeconfig {{hawkular_tmp_conf}}
-  changed_when: false
+- fail:
+    msg: This role required openshift_master_default_subdomain or openshift_master_metrics_url be set
+  when: openshift.master.metrics_public_url | default(openshift_master_metrics_public_url | default(openshift.master.default_subdomain | default(openshift_master_default_subdomain | default(none)))) is none
 
-- name: Create metrics-deployer Service Account
-  shell: >
-    echo {{ deployer_service_account | to_json | quote }} |
-    {{ openshift.common.client_binary }} create
-    -n openshift-infra
-    --config={{hawkular_tmp_conf}}
-    -f -
-  register: deployer_create_service_account
-  failed_when: "'already exists' not in deployer_create_service_account.stderr and deployer_create_service_account.rc != 0"
-  changed_when: deployer_create_service_account.rc == 0
+- name: Create temp directory for kubeconfig
+  command: mktemp -d /tmp/openshift-ansible-XXXXXX
+  register: mktemp
+  changed_when: False
 
-- name: Create metrics-deployer Secret
-  command: >
-    {{ openshift.common.client_binary }}
-    secrets new metrics-deployer
-    nothing=/dev/null
-    --config={{hawkular_tmp_conf}}
-    -n openshift-infra
-  register: deployer_create_secret
-  failed_when: "'already exists' not in deployer_create_secret.stderr and deployer_create_secret.rc !=0"
-  changed_when: deployer_create_secret.rc == 0
+- name: Record kubeconfig tmp dir
+  set_fact:
+    openshift_metrics_kubeconfig: "{{ mktemp.stdout }}/admin.kubeconfig"
 
-- name: Configure role/user permissions
+- name: Copy the admin client config(s)
   command: >
-    {{ openshift.common.admin_binary }} {{item}}
-    --config={{hawkular_tmp_conf}}
-  with_items: "{{hawkular_permission_oc_commands}}"
-  register: hawkular_perm_task
-  failed_when: "'already exists' not in hawkular_perm_task.stderr and hawkular_perm_task.rc != 0"
-  changed_when: hawkular_perm_task.rc == 0
+    cp {{ openshift_master_config_dir }}/admin.kubeconfig {{ openshift_metrics_kubeconfig }}
+  changed_when: False
+
+- name: Set hosted metrics facts
+  openshift_facts:
+    role: hosted
+    openshift_env: "{{ hostvars
+                       | oo_merge_hostvars(vars, inventory_hostname)
+                       | oo_openshift_env }}"
+    openshift_env_structures:
+    - 'openshift.hosted.metrics.*'
+
+- set_fact:
+    # Prefer the master facts over bare variables if present, prefer
+    # metrics_public_url over creating a default using default_subdomain
+    metrics_hostname: "{{ openshift.hosted.metrics.public_url
+                          | default('hawkular-metrics.' ~ (openshift.master.default_subdomain
+                          | default(openshift_master_default_subdomain )))
+                          | oo_hostname_from_url }}"
+    metrics_persistence: True
+    #"{{ openshift.hosted.metrics.storage_kind | default(none) is not none }}"
+    metrics_dynamic_vol: "{{ openshift.hosted.metrics.storage_kind | default(none) == 'dynamic' }}"
+    metrics_template_dir: "/usr/share/openshift/examples/infrastructure-templates/{{ 'origin' if deployment_type == 'origin' else 'enterprise' }}"
+    cassandra_nodes: "{{ ',CASSANDRA_NODES=' ~ openshift.hosted.metrics.cassandra_nodes if 'cassandra' in openshift.hosted.metrics else '' }}"
+    cassandra_pv_size: "{{ ',CASSANDRA_PV_SIZE=' ~ openshift.hosted.metrics.storage_volume_size if openshift.hosted.metrics.storage_volume_size | default(none) is not none else '' }}"
+    image_prefix: "{{ ',IMAGE_PREFIX=' ~ openshift.hosted.metrics.deployer_prefix if 'deployer_prefix' in openshift.hosted.metrics else '' }}"
+    image_version: "{{ ',IMAGE_VERSION=' ~ openshift.hosted.metrics.deployer_version if 'deployer_version' in openshift.hosted.metrics else '' }}"
 
-- name: Check openshift_master_default_subdomain
-  fail: 
-    msg: "Default subdomain should be defined"
-  when: openshift.master.default_subdomain is not defined
 
-- name: Create Heapster and Hawkular/Cassandra Services
+- name: Check for existing metrics pods
   shell: >
-   {{ openshift.common.client_binary }} process -f \
-   /usr/share/openshift/examples/infrastructure-templates/{{ hawkular_type }}/metrics-deployer.yaml -v \
-    HAWKULAR_METRICS_HOSTNAME=hawkular-metrics.{{ openshift.master.default_subdomain }},USE_PERSISTENT_STORAGE={{ hawkular_persistence }},METRIC_DURATION={{ openshift.hosted.metrics.duration }},METRIC_RESOLUTION={{ openshift.hosted.metrics.resolution }} \
-    | {{ openshift.common.client_binary }} create -n openshift-infra --config={{hawkular_tmp_conf}} -f -
-  register: oex_heapster_services
-  failed_when: "'already exists' not in oex_heapster_services.stderr and oex_heapster_services.rc != 0"
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    get pods -l {{ item }} | grep -q Running
+  register: metrics_pods_status
+  with_items:
+    - metrics-infra=hawkular-metrics
+    - metrics-infra=heapster
+    - metrics-infra=hawkular-cassandra
+  failed_when: false
   changed_when: false
 
-- name: Clean temporary config file
-  command: >
-    rm -rf {{hawkular_tmp_conf}}
+- name: Check for previous deployer
+  shell: >
+    {{ openshift.common.client_binary }}
+    --config={{ openshift_metrics_kubeconfig }}
+    --namespace openshift-infra
+    get pods -l metrics-infra=deployer --sort-by='{.metadata.creationTimestamp}' | tail -1 | grep metrics-deployer-
+  register: metrics_deployer_status
+  failed_when: false
   changed_when: false
 
-- name: "Wait for image pull and deployer pod"
-  shell:  "{{ openshift.common.client_binary }} get pods -n openshift-infra | grep metrics-deployer.*Completed"
-  register: result
-  until: result.rc == 0
-  retries: 60
-  delay: 10
+- name: Record current deployment status
+  set_fact:
+    greenfield: "{{ not metrics_deployer_status.rc == 0 }}"
+    failed_error: "{{ True if 'Error' in metrics_deployer_status.stdout else False }}"
+    metrics_running: "{{ metrics_pods_status.results | oo_collect(attribute='rc') == [0,0,0] }}"
+
+- name: Set deployment mode
+  set_fact:
+    deployment_mode: "{{ 'refresh' if (failed_error | bool or metrics_upgrade | bool) else 'deploy' }}"
+
+# TODO: handle non greenfield deployments in the future
+- include: install.yml
+  when: greenfield
 
+- name: Delete temp directory
+  file:
+    name: "{{ mktemp.stdout }}"
+    state: absent
+  changed_when: False
diff --git a/roles/openshift_metrics/vars/main.yaml b/roles/openshift_metrics/vars/main.yaml
index 82d9d29f7..0331bcb89 100644
--- a/roles/openshift_metrics/vars/main.yaml
+++ b/roles/openshift_metrics/vars/main.yaml
@@ -2,13 +2,13 @@ hawkular_permission_oc_commands:
     - policy add-role-to-user edit system:serviceaccount:openshift-infra:metrics-deployer -n openshift-infra
     - policy add-cluster-role-to-user cluster-admin system:serviceaccount:openshift-infra:heapster
 
-deployer_service_account:
-    apiVersion: v1
-    kind: ServiceAccount
-    metadata:
-      name: metrics-deployer
-    secrets:
-    - name: metrics-deployer
+metrics_deployer_sa:
+  apiVersion: v1
+  kind: ServiceAccount
+  metadata:
+    name: metrics-deployer
+  secrets:
+  - name: metrics-deployer
 
 
 hawkular_tmp_conf: /tmp/hawkular_admin.kubeconfig
@@ -17,3 +17,4 @@ hawkular_persistence: "{% if openshift.hosted.metrics.storage.kind != None %}tru
 
 hawkular_type: "{{ 'origin' if deployment_type == 'origin' else 'enterprise' }}"
 
+metrics_upgrade: openshift.hosted.metrics.upgrade | default(False)