NFV: Support for config-download to deploy node with kernel args

For NFV deployments, specific kernel args should be applied and the nodes should be restarted before running the NetworkDeployment. It is supported in the heat deployment via PreNetworkConfig. In the config-download mechanism, ansible steps need to be improved to handle the reboot and wait for the node. Change-Id: I43b383ad0e04b8be6c321f8c5b05e628b2520141
2018-04-11 18:38:37 +05:30 · 2018-04-11 18:38:37 +05:30 · a3e4a90636
commit a3e4a90636
parent 7f23133651
12 changed files with 168 additions and 11 deletions
--- a/common/deploy-steps.j2
+++ b/common/deploy-steps.j2
@ -714,3 +714,4 @@ outputs:
              with_sequence: start=0 end={{post_update_steps_max-1}}
              loop_control:
                loop_var: step
+      boot_param_tasks: {get_file: ../extraconfig/pre_network/boot_param_tasks.yaml}
--- a/environments/disable-config-download-environment.yaml
+++ b/environments/disable-config-download-environment.yaml
@ -4,3 +4,8 @@ resource_registry:
  OS::TripleO::Ssh::HostPubKey: ../extraconfig/tasks/ssh/host_public_key.yaml
  OS::TripleO::Ssh::KnownHostsDeployment: OS::Heat::StructuredDeployments
  OS::TripleO::DeploymentSteps: OS::Heat::StructuredDeploymentGroup
+
+  # Used only for config-download deployment, for non-config-download deployment,
+  # PreNetworkConfig (host-config-and-reboot.yaml) stack will handle it.
+  OS::TripleO::Services::BootParams: OS::Heat::None
+  OS::TripleO::Reboot::SoftwareDeployment: OS::TripleO::SoftwareDeployment
--- a/extraconfig/pre_network/boot-params-service.yaml
+++ b/extraconfig/pre_network/boot-params-service.yaml
@ -0,0 +1,102 @@
+heat_template_version: rocky
+
+description: Host config and reboot service which configures Kernel Args and reboot
+
+parameters:
+  RoleNetIpMap:
+    default: {}
+    type: json
+  ServiceData:
+    default: {}
+    description: Dictionary packing service data
+    type: json
+  ServiceNetMap:
+    default: {}
+    description: Mapping of service_name -> network name. Typically set
+                 via parameter_defaults in the resource registry.  This
+                 mapping overrides those in ServiceNetMapDefaults.
+    type: json
+  DefaultPasswords:
+    default: {}
+    type: json
+  RoleName:
+    default: ''
+    description: Role name on which the service is applied
+    type: string
+  RoleParameters:
+    default: {}
+    description: Parameters specific to the role
+    type: json
+  EndpointMap:
+    default: {}
+    description: Mapping of service endpoint -> protocol. Typically set
+                 via parameter_defaults in the resource registry.
+    type: json
+  KernelArgs:
+    default: ""
+    type: string
+    description: Kernel Args to apply to the host
+    tags:
+      - role_specific
+  TunedProfileName:
+    default: ""
+    type: string
+    description: Tuned Profile to apply to the host
+    tags:
+      - role_specific
+  IsolCpusList:
+    default: ""
+    type: string
+    description: >
+      List of logical CPU ids whic need to be isolated from the host processes.
+      This input is provided to the tuned profile cpu-partitioning to configure
+      systemd and repin interrupts (IRQ repinning).
+    tags:
+      - role_specific
+  NodeRebootWaitTimeout:
+    default: 900
+    type: number
+    description: Timeout in seconds to specify the wait time for ansible node reboot
+
+conditions:
+  is_host_config_required: {not: {equals: [{get_param: [RoleParameters, KernelArgs]}, ""]}}
+
+outputs:
+  role_data:
+    description: Role data for the Host Config Service
+    value:
+      service_name: boot_params_service
+      deploy_steps_tasks:
+        - if:
+          - is_host_config_required
+          - - name: Configure kernel args, tuned and reboot
+              when: step == 0
+              vars:
+                _KERNEL_ARGS_: {get_param: [RoleParameters, KernelArgs]}
+                _TUNED_PROFILE_NAME_: {get_param: [RoleParameters, TunedProfileName]}
+                _TUNED_CORES_: {get_param: [RoleParameters, IsolCpusList]}
+              block:
+                - import_tasks: ../boot_param_tasks.yaml
+
+                # Reboot the node
+                - name: Reboot after kernel args update
+                  shell: "sleep 2 && /sbin/shutdown -r now"
+                  async: 1
+                  poll: 0
+                  ignore_errors: true
+                  become: yes
+                  when: reboot_required is defined and reboot_required
+
+                - name:  Wait for the ssh port to be ready on the restarted node
+                  local_action: wait_for
+                  become: no
+                  args:
+                    host: "{{ inventory_hostname }}"
+                    port: 22
+                    state: started
+                    delay: 30
+                    timeout:
+                      get_param: NodeRebootWaitTimeout
+                  when: reboot_required is defined and reboot_required
+          - null
+
--- a/extraconfig/pre_network/ansible_host_config.yaml
+++ b/extraconfig/pre_network/ansible_host_config.yaml
@ -1,9 +1,12 @@
---
- name: Configuration to be applied before rebooting the node
-  connection: local
-  hosts: localhost
+    - name: Get the command line args of the node
+      command: cat /proc/cmdline
+      register: cmdline
+
+    - name: Get the active tuned profile
+      command: tuned-adm active
+      become: true
+      register: tuned_active_profile

-  tasks:
    # Kernel Args Configuration
    - block:
        - name: Ensure the kernel args ( {{ _KERNEL_ARGS_ }} ) is present as TRIPLEO_HEAT_TEMPLATE_KERNEL_ARGS
@ -19,8 +22,13 @@
            insertafter: '^TRIPLEO_HEAT_TEMPLATE_KERNEL_ARGS.*'
        - name: Generate grub config file
          command: grub2-mkconfig -o /boot/grub2/grub.cfg
+        - name: Set reboot required fact
+          set_fact:
+            reboot_required: true
      become: true
-      when: _KERNEL_ARGS_|default("") != ""
+      when:
+        - _KERNEL_ARGS_|default("") != ""
+        - _KERNEL_ARGS_|default("") not in cmdline.stdout_lines[0]

    # Tune-d Configuration
    - block:
@ -33,8 +41,20 @@

        - name: Tune-d profile activation
          shell: tuned-adm profile {{ _TUNED_PROFILE_NAME_ }}
+        - name: Set reboot required fact
+          set_fact:
+            reboot_required: true
      become: true
-      when: _TUNED_PROFILE_NAME_|default("") != ""
+      when:
+        - _TUNED_PROFILE_NAME_|default("") != ""
+        - _TUNED_PROFILE_NAME_|default("") not in tuned_active_profile.stdout_lines[0]
+
+    # Check if os-net-config has run once, if yes, no need for the below workaround
+    - find:
+        paths: /etc/sysconfig/network-scripts/
+        patterns: ifcfg-*
+        contains: "# This file is autogenerated by os-net-config"
+      register: os_net_ifcfg_files

    # Provisioning Network workaround
    # The script will be executed before os-net-config, in which case, only Provisioning network will have IP
@ -51,8 +71,12 @@
          replace: 'BOOTPROTO=none'
        when:
          - item.path | regex_replace('(^.*ifcfg-)(.*)', '\\2') != "lo"
+          # Ensure the interface information is available in the facts
+          - hostvars[inventory_hostname]['ansible_' + item.path | regex_replace('(^.*ifcfg-)(.*)', '\\2') ] is defined
          # This condition will list all the interfaces except the one with valid IP (which is Provisioning network at this stage)
          # Simpler Version - hostvars[inventory_hostname]['ansible_' + iface_name ]['ipv4'] is undefined
          - hostvars[inventory_hostname]['ansible_' + item.path | regex_replace('(^.*ifcfg-)(.*)', '\\2') ]['ipv4'] is undefined
        with_items:
          - "{{ ifcfg_files.files }}"
+      become: true
+      when: os_net_ifcfg_files.matched == 0
--- a/extraconfig/pre_network/host_config_and_reboot.yaml
+++ b/extraconfig/pre_network/host_config_and_reboot.yaml
@ -179,10 +179,20 @@ resources:
      outputs:
        - name: result
      config:
-        get_file: ansible_host_config.yaml
+        str_replace:
+          template: |
+            ---
+            - name: Configuration to be applied before rebooting the node
+              connection: local
+              hosts: localhost
+              tasks:
+            _HOST_CONFIG_TASKS_
+          params:
+            _HOST_CONFIG_TASKS_:
+              get_file: boot_param_tasks.yaml

  HostParametersDeployment:
-    type: OS::Heat::SoftwareDeployment
+    type: OS::TripleO::Reboot::SoftwareDeployment
    condition: is_host_config_required
    properties:
      name: HostParametersDeployment
@ -247,7 +257,7 @@ resources:
        /sbin/reboot

  RebootDeployment:
-    type: OS::Heat::SoftwareDeployment
+    type: OS::TripleO::Reboot::SoftwareDeployment
    depends_on: [HostParametersDeployment, DpdkVhostGroupDeployment]
    condition: is_reboot_config_required
    properties:
@ -280,7 +290,7 @@ resources:
        echo "Reboot completed"

  RebootEnsureDeployment:
-    type: OS::Heat::SoftwareDeployment
+    type: OS::TripleO::Reboot::SoftwareDeployment
    depends_on: RebootDeployment
    condition: is_reboot_config_required
    properties:
--- a/overcloud-resource-registry-puppet.j2.yaml
+++ b/overcloud-resource-registry-puppet.j2.yaml
@ -13,6 +13,8 @@ resource_registry:
  OS::TripleO::DefaultPasswords: default_passwords.yaml
  OS::TripleO::RandomString: OS::Heat::RandomString

+  OS::TripleO::Reboot::SoftwareDeployment: OS::Heat::None
+
 {% for role in roles %}
  OS::TripleO::{{role.name}}::PreNetworkConfig: OS::Heat::None
  OS::TripleO::{{role.name}}PostDeploySteps: common/post.yaml
@ -364,6 +366,8 @@ resource_registry:
  # Tempest
  OS::TripleO::Services::Tempest: OS::Heat::None

+  OS::TripleO::Services::BootParams: extraconfig/pre_network/boot-params-service.yaml
+
  # Deprecated, only defined to allow smooth transition of existing
  # stacks. Can be removed in S release.
  OS::TripleO::Tasks::UpdateWorkflow: OS::Heat::None
--- a/roles/ComputeOvsDpdk.yaml
+++ b/roles/ComputeOvsDpdk.yaml
@ -14,6 +14,7 @@
  ServicesDefault:
    - OS::TripleO::Services::Aide
    - OS::TripleO::Services::AuditD
+    - OS::TripleO::Services::BootParams
    - OS::TripleO::Services::CACerts
    - OS::TripleO::Services::CephClient
    - OS::TripleO::Services::CephExternal
--- a/roles/ComputeOvsDpdkRT.yaml
+++ b/roles/ComputeOvsDpdkRT.yaml
@ -16,6 +16,7 @@
  ServicesDefault:
    - OS::TripleO::Services::Aide
    - OS::TripleO::Services::AuditD
+    - OS::TripleO::Services::BootParams
    - OS::TripleO::Services::CACerts
    - OS::TripleO::Services::CephClient
    - OS::TripleO::Services::CephExternal
--- a/roles/ComputeRealTime.yaml
+++ b/roles/ComputeRealTime.yaml
@ -22,6 +22,7 @@
  ServicesDefault:
    - OS::TripleO::Services::Aide
    - OS::TripleO::Services::AuditD
+    - OS::TripleO::Services::BootParams
    - OS::TripleO::Services::CACerts
    - OS::TripleO::Services::CephClient
    - OS::TripleO::Services::CephExternal
--- a/roles/ComputeSriov.yaml
+++ b/roles/ComputeSriov.yaml
@ -13,6 +13,7 @@
  ServicesDefault:
    - OS::TripleO::Services::Aide
    - OS::TripleO::Services::AuditD
+    - OS::TripleO::Services::BootParams
    - OS::TripleO::Services::CACerts
    - OS::TripleO::Services::CephClient
    - OS::TripleO::Services::CephExternal
--- a/roles/ComputeSriovRT.yaml
+++ b/roles/ComputeSriovRT.yaml
@ -16,6 +16,7 @@
  ServicesDefault:
    - OS::TripleO::Services::Aide
    - OS::TripleO::Services::AuditD
+    - OS::TripleO::Services::BootParams
    - OS::TripleO::Services::CACerts
    - OS::TripleO::Services::CephClient
    - OS::TripleO::Services::CephExternal
--- a/tools/yaml-validate.py
+++ b/tools/yaml-validate.py
@ -240,6 +240,10 @@ WORKFLOW_TASKS_EXCLUSIONS = [
 ]


+ANSIBLE_TASKS_YAMLS = [
+    './extraconfig/pre_network/boot_param_tasks.yaml'
+]
+
 def exit_usage():
    print('Usage %s <yaml file or directory>' % sys.argv[0])
    sys.exit(1)
@ -1034,6 +1038,8 @@ def validate(filename, param_map):
            retval |= validate_nic_config_file(filename, tpl)

    except Exception:
+        if filename in ANSIBLE_TASKS_YAMLS:
+            return 0
        print(traceback.format_exc())
        return 1
    # yaml is OK, now walk the parameters and output a warning for unused ones