[Linux] Fix system-probe enablement conditions

README.md

@@ -163,7 +163,7 @@ The following variables are available for live processes:
 
 #### System probe
 
-The system probe is configured under the `network_config` variable. Any variables nested underneath are written to the `system-probe.yaml`.
+The system probe is configured under the `system_probe_config` variable. Any variables nested underneath are written to the `system-probe.yaml`, in the `system_probe_config` section.
 
 [Network Performance Monitoring][7] (NPM) is configured under the `network_config` variable.  Any variables nested underneath are written to the `system-probe.yaml`, in the `network_config` section.
 
@@ -183,13 +183,15 @@ network_config:
   enabled: true
 ```
 
-Once modification is complete, follow the steps below:
+**Note**: This configuration works with Agent 6.24.1+ and 7.24.1+. For older Agent versions, refer to [the public documentation][8] on how to enable system-probe.
+
+On Linux, once this modification is complete, follow the steps below if you installed an Agent version older than 6.18.0 or 7.18.0:
 
 1. Start the system-probe: `sudo service datadog-agent-sysprobe start` **Note**: If the service wrapper is not available on your system, run this command instead: `sudo initctl start datadog-agent-sysprobe`.
-2. [Restart the Agent][8]: `sudo service datadog-agent restart`.
+2. [Restart the Agent][9]: `sudo service datadog-agent restart`.
 3. Enable the system-probe to start on boot: `sudo service enable datadog-agent-sysprobe`.
 
-For manual setup, refer to the [NPM][9] documentation.
+For manual setup, refer to the [NPM][8] documentation.
 
 #### Agent v5
 
@@ -334,7 +336,7 @@ To downgrade to a prior version of the Agent:
 
 Below are some sample playbooks to assist you with using the Datadog Ansible role.
 
-The following example sends data to Datadog US (default), enables logs, and configures a few checks.
+The following example sends data to Datadog US (default), enables logs, NPM and configures a few checks.
 
 ```yml
 - hosts: servers
@@ -403,7 +405,7 @@ The following example sends data to Datadog US (default), enables logs, and conf
         version: 1.11.0
       datadog-postgres:
         action: remove
-    system_probe_config:
+    network_config:
       enabled: true
 ```
 
@@ -530,6 +532,6 @@ For more details, see [Critical Bug in Uninstaller for Datadog Agent 6.14.0 and
 [5]: https://github.com/DataDog/integrations-core
 [6]: https://docs.datadoghq.com/infrastructure/process/
 [7]: https://docs.datadoghq.com/network_performance_monitoring/
-[8]: https://docs.datadoghq.com/agent/guide/agent-commands/#restart-the-agent
-[9]: https://docs.datadoghq.com/network_performance_monitoring/installation/?tab=agent#setup
+[8]: https://docs.datadoghq.com/network_performance_monitoring/installation/?tab=agent#setup
+[9]: https://docs.datadoghq.com/agent/guide/agent-commands/#restart-the-agent
 [10]: https://app.datadoghq.com/help/agent_fix

ci_test/install_agent_6.yaml

@@ -18,6 +18,10 @@
         env: dev
       trace.concentrator:
         extra_aggregators: version
+    system_probe_config:
+      sysprobe_socket: /opt/datadog-agent/run/sysprobe.sock
+    network_config:
+      enabled: true
     datadog_checks:
       process:
         init_config:

ci_test/install_agent_7.yaml

@@ -18,6 +18,10 @@
         env: dev
       trace.concentrator:
         extra_aggregators: version
+    system_probe_config:
+      sysprobe_socket: /opt/datadog-agent/run/sysprobe.sock
+    network_config:
+      enabled: true
     datadog_checks:
       process:
         init_config:

defaults/main.yml

@@ -2,11 +2,9 @@
 role_version: 4.7.1
 
 # default system-probe.yaml options
-system_probe_config:
-  enabled: false
+system_probe_config: {}
 
-network_config:
-  enabled: false
+network_config: {}
 
 # define if the datadog-agent services should be enabled
 datadog_enabled: yes

handlers/main.yml

@@ -1,5 +1,11 @@
 ---
 
+- name: restart datadog-agent-sysprobe
+  service:
+    name: datadog-agent-sysprobe
+    state: restarted
+  when: datadog_enabled and datadog_sysprobe_enabled and not ansible_check_mode and not ansible_facts.os_family == "Windows"
+
 - name: restart datadog-agent
   service:
     name: datadog-agent

manual_tests/test_6_full.yml

@@ -6,13 +6,14 @@
     datadog_api_key: "123456"
     datadog_agent_allow_downgrade: true
     system_probe_config:
-      enabled: true
       source_excludes:
         "*":
           - 8301
       dest_excludes:
         "*":
           - 8301
+    network_config:
+      enabled: true
     datadog_config:
       tags: "mytag0, mytag1"
       log_level: INFO

manual_tests/test_7_full.yml

@@ -6,13 +6,14 @@
     datadog_api_key: "123456"
     datadog_agent_allow_downgrade: true
     system_probe_config:
-      enabled: true
       source_excludes:
         "*":
           - 8301
       dest_excludes:
         "*":
           - 8301
+    network_config:
+      enabled: true
     datadog_config:
       tags: "mytag0, mytag1"
       log_level: INFO

tasks/agent-linux.yml

@@ -1,8 +1,19 @@
 ---
-- name: populate service facts
+- name: Populate service facts
   service_facts:
 
-- name: add "{{ datadog_user }}" user to additional groups
+- name: Set before 6/7.24.1 flag
+  set_fact:
+    datadog_before_7241: "{{ datadog_major is defined and datadog_minor is defined and datadog_bugfix is defined
+      and datadog_major | int < 8
+      and (datadog_minor | int < 24 or (datadog_minor | int == 24 and datadog_bugfix | int < 1)) }}"
+
+- name: Set before 6/7.18.0 flag
+  set_fact:
+    datadog_before_7180: "{{ datadog_major is defined and datadog_minor is defined
+      and datadog_major | int < 8 and datadog_minor | int < 18 }}"
+
+- name: Add "{{ datadog_user }}" user to additional groups
   user: name="{{ datadog_user }}" groups="{{ datadog_additional_groups }}" append=yes
   when: datadog_additional_groups | default([], true) | length > 0
   notify: restart datadog-agent
@@ -78,40 +89,55 @@
     mode: 0640
     owner: "root"
     group: "{{ datadog_group }}"
+  notify:
+    "{% if datadog_before_7180 %}restart datadog-agent-sysprobe{% else %}restart datadog-agent{% endif %}"
 
-- name: Ensure datadog-agent is running
-  service:
-    name: datadog-agent
-    state: started
-    enabled: yes
-  when: not datadog_skip_running_check and datadog_enabled and not ansible_check_mode
-
-- name: set system probe installed
+- name: Set system probe installed
   set_fact:
     datadog_sysprobe_installed: "{{ ansible_facts.services['datadog-agent-sysprobe'] is defined
       or ansible_facts.services['datadog-agent-sysprobe.service'] is defined }}"
   when: not datadog_skip_running_check
 
-- name: set system probe enabled
+# Before 6/7.24.1, system_probe_config controls the system-probe service
+# datadog_minor is only defined when a specific Agent version is given
+# (see tasks/parse-version.yml)
+- name: Set system probe enabled (before 6/7.24.1)
   set_fact:
     datadog_sysprobe_enabled: "{{ system_probe_config is defined
+      and 'enabled' in (system_probe_config | default({}, true))
       and system_probe_config['enabled']
       and datadog_sysprobe_installed }}"
   when: not datadog_skip_running_check
+    and datadog_before_7241
 
-- name: Ensure datadog-agent-sysprobe is running if enabled and installed
+# Since 6/7.24.1, setting enabled: true in network_config is enough to start the system-probe service:
+# https://docs.datadoghq.com/network_monitoring/performance/setup/?tab=agent#setup
+- name: Set system probe enabled (since 6/7.24.1)
+  set_fact:
+    datadog_sysprobe_enabled: "{{
+      ((system_probe_config is defined
+      and 'enabled' in (system_probe_config | default({}, true))
+      and system_probe_config['enabled'])
+      or (network_config is defined
+      and 'enabled' in (network_config | default({}, true))
+      and network_config['enabled']))
+      and datadog_sysprobe_installed }}"
+  when: not datadog_skip_running_check
+    and (not datadog_before_7241)
+
+- name: Ensure datadog-agent is running
   service:
-    name: datadog-agent-sysprobe
+    name: datadog-agent
     state: started
     enabled: yes
-  when: not datadog_skip_running_check and datadog_enabled and not ansible_check_mode and datadog_sysprobe_enabled
+  when: not datadog_skip_running_check and datadog_enabled and not ansible_check_mode
 
-- name: Ensure datadog-agent-sysprobe is stopped if disabled or not installed
+- name: Ensure datadog-agent-sysprobe is running if enabled and installed
   service:
     name: datadog-agent-sysprobe
-    state: stopped
-    enabled: no
-  when: not datadog_skip_running_check and (not datadog_enabled or not datadog_sysprobe_enabled) and datadog_sysprobe_installed
+    state: started
+    enabled: yes
+  when: not datadog_skip_running_check and datadog_enabled and not ansible_check_mode and datadog_sysprobe_enabled
 
 - name: Ensure datadog-agent, datadog-agent-process and datadog-agent-trace are not running
   service:
@@ -124,6 +150,18 @@
     - datadog-agent-process
     - datadog-agent-trace
 
+# Stop system-probe manually on Agent versions < 6/7.18, as it was not tied
+# to the main Agent service: https://github.com/DataDog/datadog-agent/pull/4883
+- name: Ensure datadog-agent-sysprobe is stopped if disabled or not installed (before 6/7.18.0)
+  service:
+    name: datadog-agent-sysprobe
+    state: stopped
+    enabled: no
+  when: not datadog_skip_running_check
+    and (not datadog_enabled or not datadog_sysprobe_enabled)
+    and datadog_before_7180
+    and datadog_sysprobe_installed
+
 - name: Ensure datadog-agent-security is not running
   service:
     name: datadog-agent-security

templates/system-probe.yaml.j2

@@ -1,6 +1,6 @@
 # Managed by Ansible
 
-{% if system_probe_config is defined and system_probe_config|length > 0 -%}
+{% if system_probe_config is defined and system_probe_config | default({}, true) | length > 0 -%}
 system_probe_config:
 {# The "first" option is only supported by jinja 2.10+
   which is not present on older systems (CentOS 7, Debian 8, etc.)