Use ansible for nodes validation

The old all nodes validation used a bash script to run some basic ping tests after the network setup. It used to be a software config but eventually got baked into the deployment framework. This patch switches to the ansible role implementation and cleans up the old references to the old heat resource. Change-Id: Ia7f055d2c636f950c3fe6d8611834c4ab290f31a Depends-On: https://review.opendev.org/#/c/747466/
11 months ago · 78ca1fe143
18 changed files with 8 additions and 135 deletions
--- a/all-nodes-validation.yaml
+++ b/all-nodes-validation.yaml
@ -1,45 +0,0 @@
-heat_template_version: rocky
-
-description: >
-  Software Config to drive validations that occur on all nodes.
-  Note, you need the heat-config-script element built into your
-  images, due to the script group below.
-
-parameters:
-  PingTestIps:
-    default: ''
-    description: A string containing a space separated list of IP addresses used to ping test each available network interface.
-    type: string
-  ValidateControllersIcmp:
-    default: true
-    description: Validation to ensure that all controllers can be reached with ICMP
-    type: boolean
-  ValidateGatewaysIcmp:
-    default: true
-    description: Validation to ensure that all gateways can be reached with ICMP
-    type: boolean
-  ValidateFqdn:
-    default: false
-    description: Optional validation to ensure FQDN as set by Nova matches the name set in /etc/hosts.
-    type: boolean
-
-resources:
-  AllNodesValidationsImpl:
-    type: OS::Heat::SoftwareConfig
-    properties:
-      group: script
-      inputs:
-        - name: ping_test_ips
-          default: {get_param: PingTestIps}
-        - name: validate_fqdn
-          default: {get_param: ValidateFqdn}
-        - name: validate_controllers_icmp
-          default: {get_param: ValidateControllersIcmp}
-        - name: validate_gateways_icmp
-          default: {get_param: ValidateGatewaysIcmp}
-      config: {get_file: ./validation-scripts/all-nodes.sh}
-
-outputs:
-  OS::stack_id:
-    description: The ID of the AllNodesValidationsImpl resource.
-    value: {get_resource: AllNodesValidationsImpl}
--- a/ci/common/all-nodes-validation-disabled.yaml
+++ b/ci/common/all-nodes-validation-disabled.yaml
@ -1,37 +0,0 @@
-heat_template_version: rocky
-
-description: >
-  Software Config to drive validations that occur on all nodes.
-  Note, you need the heat-config-script element built into your
-  images, due to the script group below.
-
-  This implementation of the validations is a noop that always reports success.
-
-parameters:
-  PingTestIps:
-    default: ''
-    description: A string containing a space separated list of IP addresses used to ping test each available network interface.
-    type: string
-  ValidateFqdn:
-    default: false
-    description: Optional validation to ensure FQDN as set by Nova matches the name set in /etc/hosts.
-    type: boolean
-
-resources:
-  AllNodesValidationsImpl:
-    type: OS::Heat::SoftwareConfig
-    properties:
-      group: script
-      inputs:
-        - name: ping_test_ips
-          default: {get_param: PingTestIps}
-        - name: validate_fqdn
-          default: {get_param: ValidateFqdn}
-      config: |
-        #!/bin/bash
-        exit 0
-
-outputs:
-  OS::stack_id:
-    description: The ID of the AllNodesValidationsImpl resource.
-    value: {get_resource: AllNodesValidationsImpl}
--- a/ci/environments/multinode-containers.yaml
+++ b/ci/environments/multinode-containers.yaml
@ -1,9 +1,6 @@
 resource_registry:
  OS::TripleO::Controller::Net::SoftwareConfig: ../common/net-config-multinode.yaml
  OS::TripleO::Compute::Net::SoftwareConfig: ../common/net-config-multinode.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::OVNDBs: ../../deployment/ovn/ovn-dbs-pacemaker-puppet.yaml

 parameter_defaults:
--- a/ci/environments/scenario000-multinode-containers.yaml
+++ b/ci/environments/scenario000-multinode-containers.yaml
@ -1,9 +1,6 @@
 resource_registry:
  OS::TripleO::Controller::Net::SoftwareConfig: ../common/net-config-multinode.yaml
  OS::TripleO::Compute::Net::SoftwareConfig: ../common/net-config-multinode.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml

 parameter_defaults:
  ControllerServices:
--- a/ci/environments/scenario000-standalone.yaml
+++ b/ci/environments/scenario000-standalone.yaml
@ -1,8 +1,5 @@
 # This scenario only install keystone and the required services to run it.
 resource_registry:
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::AodhApi: OS::Heat::None
  OS::TripleO::Services::AodhEvaluator: OS::Heat::None
  OS::TripleO::Services::AodhListener: OS::Heat::None
--- a/ci/environments/scenario001-multinode-containers.yaml
+++ b/ci/environments/scenario001-multinode-containers.yaml
@ -28,9 +28,6 @@ resource_registry:
  OS::TripleO::Services::HeatApiCfn: ../../deployment/heat/heat-api-cfn-container-puppet.yaml
  OS::TripleO::Services::HeatApiCloudwatch: ../../deployment/heat/heat-api-cloudwatch-disabled-puppet.yaml
  OS::TripleO::Services::HeatEngine: ../../deployment/heat/heat-engine-container-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml

 parameter_defaults:
  ControllerServices:
--- a/ci/environments/scenario001-standalone.yaml
+++ b/ci/environments/scenario001-standalone.yaml
@ -25,9 +25,6 @@ resource_registry:
  OS::TripleO::Services::Redis: ../../deployment/database/redis-pacemaker-puppet.yaml
  OS::TripleO::Services::CinderBackup: ../../deployment/cinder/cinder-backup-pacemaker-puppet.yaml
  OS::TripleO::Services::CinderVolume: ../../deployment/cinder/cinder-volume-pacemaker-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::SwiftProxy: OS::Heat::None
  OS::TripleO::Services::SwiftStorage: OS::Heat::None
  OS::TripleO::Services::SwiftRingBuilder: OS::Heat::None
--- a/ci/environments/scenario002-standalone.yaml
+++ b/ci/environments/scenario002-standalone.yaml
@ -21,9 +21,6 @@ resource_registry:
  OS::TripleO::Services::CinderBackup: ../../deployment/cinder/cinder-backup-pacemaker-puppet.yaml
  OS::TripleO::Services::CinderVolume: ../../deployment/cinder/cinder-volume-pacemaker-puppet.yaml
  OS::TripleO::Services::BarbicanBackendSimpleCrypto: ../../deployment/barbican/barbican-backend-simple-crypto-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  # NOTE(mwhahaha): cheat and use the horizon "service" for our test service
  # since disable horizon. This allows us to not have to keep some test service
  # in the defaults just for this case.
--- a/ci/environments/scenario003-standalone.yaml
+++ b/ci/environments/scenario003-standalone.yaml
@ -18,9 +18,6 @@ resource_registry:
  OS::TripleO::Services::DesignateWorker: ../../deployment/experimental/designate/designate-worker-container-puppet.yaml
  OS::TripleO::Services::DesignateMDNS: ../../deployment/experimental/designate/designate-mdns-container-puppet.yaml
  OS::TripleO::Services::Redis: ../../deployment/database/redis-container-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml

 parameter_defaults:
  Debug: true
--- a/ci/environments/scenario004-standalone.yaml
+++ b/ci/environments/scenario004-standalone.yaml
@ -27,9 +27,6 @@ resource_registry:
  OS::TripleO::Services::Clustercheck: ../../deployment/pacemaker/clustercheck-container-puppet.yaml
  OS::TripleO::Services::MySQL: ../../deployment/database/mysql-pacemaker-puppet.yaml
  OS::TripleO::Services::LoginDefs: ../../deployment/login-defs/login-defs-baremetal.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml

 parameter_defaults:
  ManagePolling: true
--- a/ci/environments/scenario007-multinode-containers.yaml
+++ b/ci/environments/scenario007-multinode-containers.yaml
@ -10,9 +10,6 @@ resource_registry:
  OS::TripleO::Services::NeutronL3Agent: ../../deployment/neutron/neutron-l3-container-puppet.yaml
  OS::TripleO::Services::ComputeNeutronOvsAgent: ../../deployment/neutron/neutron-ovs-agent-container-puppet.yaml

-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::NovaMigrationTarget: OS::Heat::None
  OS::TripleO::Services::SwiftProxy: OS::Heat::None
  OS::TripleO::Services::SwiftStorage: OS::Heat::None
--- a/ci/environments/scenario007-standalone.yaml
+++ b/ci/environments/scenario007-standalone.yaml
@ -6,9 +6,6 @@ resource_registry:
  OS::TripleO::Services::NeutronOvsAgent: ../../deployment/neutron/neutron-ovs-agent-container-puppet.yaml
  OS::TripleO::Services::NeutronDhcpAgent: ../../deployment/neutron/neutron-dhcp-container-puppet.yaml
  OS::TripleO::Services::NeutronL3Agent: ../../deployment/neutron/neutron-l3-container-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../../ci/common/all-nodes-validation-disabled.yaml

  OS::TripleO::Services::BlockStorageCinderVolume: OS::Heat::None
  OS::TripleO::Services::CinderVolume: OS::Heat::None
--- a/ci/environments/scenario010-multinode-containers.yaml
+++ b/ci/environments/scenario010-multinode-containers.yaml
@ -10,7 +10,6 @@ resource_registry:
  OS::TripleO::Services::PacemakerRemote: ../../deployment/pacemaker/pacemaker-remote-baremetal-puppet.yaml
  OS::TripleO::Services::Clustercheck: ../../deployment/pacemaker/clustercheck-container-puppet.yaml
  OS::TripleO::Services::MySQL: ../../deployment/database/mysql-pacemaker-puppet.yaml
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::OctaviaApi: ../../deployment/octavia/octavia-api-container-puppet.yaml
  OS::TripleO::Services::OctaviaHousekeeping: ../../deployment/octavia/octavia-housekeeping-container-puppet.yaml
  OS::TripleO::Services::OctaviaHealthManager: ../../deployment/octavia/octavia-health-manager-container-puppet.yaml
--- a/ci/environments/scenario010-standalone.yaml
+++ b/ci/environments/scenario010-standalone.yaml
@ -5,9 +5,6 @@ resource_registry:
  OS::TripleO::Services::CephClient: ../../deployment/ceph-ansible/ceph-client.yaml
  OS::TripleO::Services::OsloMessagingRpc: ../../deployment/rabbitmq/rabbitmq-messaging-rpc-pacemaker-puppet.yaml
  OS::TripleO::Services::OsloMessagingNotify: ../../deployment/rabbitmq/rabbitmq-messaging-notify-shared-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::OctaviaApi: ../../deployment/octavia/octavia-api-container-puppet.yaml
  OS::TripleO::Services::OctaviaHousekeeping: ../../deployment/octavia/octavia-housekeeping-container-puppet.yaml
  OS::TripleO::Services::OctaviaHealthManager: ../../deployment/octavia/octavia-health-manager-container-puppet.yaml
--- a/ci/environments/scenario012-standalone.yaml
+++ b/ci/environments/scenario012-standalone.yaml
@ -6,9 +6,6 @@ resource_registry:
  OS::TripleO::Services::IronicApi: ../../deployment/ironic/ironic-api-container-puppet.yaml
  OS::TripleO::Services::IronicConductor: ../../deployment/ironic/ironic-conductor-container-puppet.yaml
  OS::TripleO::Services::IronicPxe: ../../deployment/ironic/ironic-pxe-container-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::NodeExtraConfigPost: ../common/ironic_standalone_post.yaml
  OS::TripleO::Services::CinderVolume: ../../deployment/cinder/cinder-volume-container-puppet.yaml
  OS::TripleO::Services::NeutronCorePlugin: OS::TripleO::Services::NeutronCorePluginML2Ansible
--- a/ci/environments/scenario013-standalone.yaml
+++ b/ci/environments/scenario013-standalone.yaml
@ -2,10 +2,6 @@
 resource_registry:
  OS::TripleO::Services::Keepalived: OS::Heat::None
  OS::TripleO::Services::OsloMessagingRpc: ../../deployment/rabbitmq/rabbitmq-messaging-rpc-container-puppet.yaml
-  OS::TripleO::Services::OsloMessagingNotify: ../../deployment/rabbitmq/rabbitmq-messaging-notify-shared-puppet.yaml
-  # Some infra instances don't pass the ping test but are otherwise working.
-  # Since the OVB jobs also test this functionality we can shut it off here.
-  OS::TripleO::AllNodes::Validation: ../common/all-nodes-validation-disabled.yaml
  OS::TripleO::Services::OctaviaApi: ../../deployment/octavia/octavia-api-container-puppet.yaml
  OS::TripleO::Services::OctaviaHousekeeping: ../../deployment/octavia/octavia-housekeeping-container-puppet.yaml
  OS::TripleO::Services::OctaviaHealthManager: ../../deployment/octavia/octavia-health-manager-container-puppet.yaml
--- a/common/deploy-steps.j2
+++ b/common/deploy-steps.j2
@ -411,7 +411,6 @@ outputs:
      deploy_steps_tasks_step_0: {get_file: deploy-steps-tasks-step-0.yaml}
      common_deploy_steps_tasks_step_1: {get_file: deploy-steps-tasks-step-1.yaml}
      container_puppet_script: {get_file: ./container-puppet.sh}
-      all_nodes_validation_script.sh : {get_file: ../validation-scripts/all-nodes.sh}
      deploy-artifacts.sh : {get_file: ../puppet/deploy-artifacts.sh}
      generate-config-tasks: {get_file: generate-config-tasks.yaml}
      host-container-puppet-tasks: {get_file: host-container-puppet-tasks.yaml}
@ -575,14 +574,14 @@ outputs:
                    tripleo_network_config_async_poll: "{{ async_poll | default(3) }}"
                  when:
                    - NetworkConfig_stat.stat.exists
-                - name: AllNodesValidationConfig
-                  script: all_nodes_validation_script.sh
-                  environment:
-                    validate_controllers_icmp: "{{ validate_controllers_icmp }}"
-                    validate_gateways_icmp: "{{ validate_gateways_icmp }}"
-                    validate_fqdn: "{{ validate_fqdn }}"
-                    ping_test_ips: "{{ ping_test_ips | to_json }}"
-                    tripleo_role_name: "{{ tripleo_role_name }}"
+                - name: Basic Network Validation
+                  include_role:
+                    name: tripleo_nodes_validation
+                  vars:
+                    tripleo_nodes_validation_validate_controllers_icmp: "{{ validate_controllers_icmp }}"
+                    tripleo_nodes_validation_validate_gateways_icmp: "{{ validate_gateways_icmp }}"
+                    tripleo_nodes_validation_validate_fqdn: "{{ validate_fqdn }}"
+                    tripleo_nodes_validation_ping_test_ips: "{{ ping_test_ips.get(tripleo_role_name).split(' ') | list | unique }}"
              tags:
                - overcloud
                - pre_deploy_steps
--- a/overcloud-resource-registry-puppet.j2.yaml
+++ b/overcloud-resource-registry-puppet.j2.yaml
@ -92,9 +92,6 @@ resource_registry:
  # Service Endpoint Mappings
  OS::TripleO::EndpointMap: network/endpoints/endpoint_map.yaml

-  # validation resources
-  OS::TripleO::AllNodes::Validation: all-nodes-validation.yaml
-
  OS::TripleO::DeployedServerEnvironment: OS::Heat::None

  OS::TripleO::DeploymentSteps: OS::Heat::None