Set env vars for service instead of machine (#3930)

* Set env vars for service instead of machine

* Fix copy and pasted typo

* Remove unused variable

* Update chocolatey to use env vars at the service scope

* Fix check for "SPLUNK_ACCESS_TOKEN"

* Handle upgrade in chocolatey

* Get-Package not available on all Windows versions

* Add messages to track stage on chocolatey install

* Adding more messages to chocolatey script

* Fix service on the registry but no env

* Restore $regKey on install.ps1

* Add helper allow running the collector manually on Windows

* Add breaking change to CHANGELOG.md

* Add -config_path switch to install.ps1

* Fix pipeline collector for Windows ZC IIS test

* MSI to set SPLUNK_CONFIG at the service scope

* Minor doc updates

* Update ansible tests for Windows env vars

* Correct type of *_ENABLE_PROFILING value

* ansible: Windows deployment updates

* fix yamllint issues

* fix yamllint

* Output vagrant.err in case of failure

* Fix splunk_trace_url check

* Always try to display the vagrant errors

* Check if event log has some info about failure

* Remove event log steps: running on mac not Windows

* Launch vagrant

* Init vagrant

* Drop launch vagrant step

* Fix ansible Windows

* Fix yamllint issues

* Remove vagrant init

* Updates to pass local run of molecule tests

* Update chef deployment

* Proper data type for multi_string registry entry

* Fix var name typo

* Fix order of collector environment variables

* Experiment custom_vars test

* Pay attention to conditional variables

* First try with Puppet

* Fix syntax for Puppet

* Back to curly brackets

* Try extension fixes

* Fix conditional

* Remove to_s

* Use empty instead of is_empty

* Fix re-declaration of registry_key

* Fix puppet lint issue

* Do not re-use registry_key

* Use array instead of multi_string

* Ensure side effects

* Update test to use REG_MULTI_SZ

* Immutability?

* Fix lint error

* Fix test

* Removing name re-use

* Update CHANGELOG and README for Puppet

* Updates to mass deployments docs

* Update last version deploying with machine wide env vars

* Sort service environment variables set via PowerShell

* Improve handling of collector restart in Ansible

* Remove extra double-quotes in SPLUNK_CONFIG for Ansible

* Fix Ansible verify SPLUNK_CONFIG strings

* Fix Windows testing checking for legacy versions of the collector

* Update CHANGELOG.md

Co-authored-by: Ryan Fitzpatrick <[email protected]>

* Update last version still setting machine wide env vars

---------

Co-authored-by: Ryan Fitzpatrick <[email protected]>

Author: pjanotti

.github/workflows/scripts/win-test-services.ps1

@@ -10,10 +10,25 @@ param (
 $ErrorActionPreference = 'Stop'
 Set-PSDebug -Trace 1
 
-function check_regkey([string]$name, [string]$value) {
-    $actual = Get-ItemPropertyValue -PATH "HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment" -name "$name"
-    if ( "$value" -ne "$actual" ) {
-        throw "Environment variable $name is not properly set. Found: '$actual', Expected '$value'"
+function check_collector_svc_environment([hashtable]$expected_env_vars) {
+    $actual_env_vars = @{}
+    try {
+        $env_array = Get-ItemPropertyValue -Path "HKLM:\SYSTEM\CurrentControlSet\Services\splunk-otel-collector" -Name "Environment"
+        foreach ($entry in $env_array) {
+            $key, $value = $entry.Split("=", 2)
+            $actual_env_vars.Add($key, $value)
+        }
+    } catch {
+        Write-Host "Assuming an old version of the collector with environment variables at the machine scope"
+        $actual_env_vars = [Environment]::GetEnvironmentVariables("Machine")<#Do this if a terminating exception happens#>
+    }
+
+    foreach ($key in $expected_env_vars.Keys) {
+        $expected_value = $expected_env_vars[$key]
+        $actual_value = $actual_env_vars[$key]
+        if ($expected_value -ne $actual_value) {
+            throw "Environment variable $key is not properly set. Found: '$actual_value', Expected '$expected_value'"
+        }
     }
 }
 
@@ -24,16 +39,18 @@ function service_running([string]$name) {
 $api_url = "https://api.${realm}.signalfx.com"
 $ingest_url = "https://ingest.${realm}.signalfx.com"
 
-check_regkey -name "SPLUNK_CONFIG" -value "${env:PROGRAMDATA}\Splunk\OpenTelemetry Collector\${mode}_config.yaml"
-check_regkey -name "SPLUNK_ACCESS_TOKEN" -value "$access_token"
-check_regkey -name "SPLUNK_REALM" -value "$realm"
-check_regkey -name "SPLUNK_API_URL" -value "$api_url"
-check_regkey -name "SPLUNK_INGEST_URL" -value "$ingest_url"
-check_regkey -name "SPLUNK_TRACE_URL" -value "${ingest_url}/v2/trace"
-check_regkey -name "SPLUNK_HEC_URL" -value "${ingest_url}/v1/log"
-check_regkey -name "SPLUNK_HEC_TOKEN" -value "$access_token"
-check_regkey -name "SPLUNK_BUNDLE_DIR" -value "${env:PROGRAMFILES}\Splunk\OpenTelemetry Collector\agent-bundle"
-check_regkey -name "SPLUNK_MEMORY_TOTAL_MIB" -value  "$memory"
+check_collector_svc_environment @{
+  "SPLUNK_CONFIG"           = "${env:PROGRAMDATA}\Splunk\OpenTelemetry Collector\${mode}_config.yaml";
+  "SPLUNK_ACCESS_TOKEN"     = "$access_token";
+  "SPLUNK_REALM"            = "$realm";
+  "SPLUNK_API_URL"          = "$api_url";
+  "SPLUNK_INGEST_URL"       = "$ingest_url";
+  "SPLUNK_TRACE_URL"        = "${ingest_url}/v2/trace";
+  "SPLUNK_HEC_URL"          = "${ingest_url}/v1/log";
+  "SPLUNK_HEC_TOKEN"        = "$access_token";
+  "SPLUNK_BUNDLE_DIR"       = "${env:PROGRAMFILES}\Splunk\OpenTelemetry Collector\agent-bundle";
+  "SPLUNK_MEMORY_TOTAL_MIB" = "$memory";
+}
 
 if ((service_running -name "splunk-otel-collector")) {
     write-host "splunk-otel-collector service is running."

CHANGELOG.md

@@ -2,6 +2,10 @@
 
 ## Unreleased
 
+### 🛑 Breaking changes 🛑
+
+- (Splunk) On Windows the `SPLUNK_*` environment variables were moved from the machine scope to the collector service scope. This avoids collisions with other agents and instrumentation. If any of these environment variables are required by your apps, please adopt them directly. ([#3930](https://github.com/signalfx/splunk-otel-collector/pull/3930))
+
 ## v0.92.0
 
 This Splunk OpenTelemetry Collector release includes changes from the [opentelemetry-collector v0.92.0](https://github.com/open-telemetry/opentelemetry-collector/releases/tag/v0.92.0) and the [opentelemetry-collector-contrib v0.92.0](https://github.com/open-telemetry/opentelemetry-collector-contrib/releases/tag/v0.92.0) releases where appropriate.
@@ -86,6 +90,7 @@ This Splunk OpenTelemetry Collector release includes changes from the [opentelem
 - (Core) `otlpexporter`: remove dependency of otlphttpreceiver on otlpexporter ([#6454](https://github.com/open-telemetry/opentelemetry-collector/issues/6454))
 
 ## v0.91.3
+
 - (Splunk) Properly sign and associate changelog to release.  This should be otherwise identical to v0.91.2
 
 ## v0.91.2
@@ -113,7 +118,6 @@ This Splunk OpenTelemetry Collector release includes changes from the [opentelem
 - (Splunk) Adopt `awss3` exporter ([#4117](https://github.com/signalfx/splunk-otel-collector/pull/4117))
 - (Splunk) Convert loglevel to verbosity on logging exporter ([#4097](https://github.com/signalfx/splunk-otel-collector/pull/4097))
 
-
 ## v0.91.1
 
 ### 💡 Enhancements 💡
@@ -301,7 +305,6 @@ This Splunk OpenTelemetry Collector release includes changes from the [opentelem
 - (Contrib) `zipkinreceiver`: Return BadRequest in case of permanent errors ([#4335](https://github.com/open-telemetry/opentelemetry-collector-contrib/issues/4335))
 - (Core) `exporterhelper`: fix bug with queue size and capacity metrics ([#8682](https://github.com/open-telemetry/opentelemetry-collector/issues/8682))
 
-
 ## v0.88.0
 
 This Splunk OpenTelemetry Collector release includes changes from the [opentelemetry-collector v0.88.0](https://github.com/open-telemetry/opentelemetry-collector/releases/tag/v0.88.0) and the [opentelemetry-collector-contrib v0.88.0](https://github.com/open-telemetry/opentelemetry-collector-contrib/releases/tag/v0.88.0) releases where appropriate.

deployments/ansible/CHANGELOG.md

@@ -2,6 +2,9 @@
 
 ## unreleased
 
+- On Windows the `SPLUNK_*` environment variables were moved from the machine scope to the collector service scope.
+  It is possible that some instrumentations are relying on the machine-wide environment variables set by the installation. ([#3930](https://github.com/signalfx/splunk-otel-collector/pull/3930))
+
 ## ansible-v0.24.0
 
 ### 🚩 Deprecations 🚩

deployments/ansible/README.md

@@ -25,6 +25,10 @@ Currently, the following Windows versions are supported:
 - Windows Server 2019 64-bit
 - Windows Server 2022 64-bit
 
+On Windows, the collector is installed as a Windows service and its environment
+variables are set at the service scope, i.e.: they are only available to the
+collector service and not to the entire machine.
+
 Ansible requires PowerShell 3.0 or newer and at least .NET 4.0 to be installed on Windows host.
 A WinRM listener should be created and activeted. 
 For setting up Windows Host refer [Ansible Docs](https://docs.ansible.com/ansible/latest/user_guide/windows_setup.html).

deployments/ansible/molecule/custom_vars/windows-verify.yml

@@ -4,7 +4,7 @@
   gather_facts: true
   become: no
   vars:
-    reg_values:
+    collector_reg_values:
       SPLUNK_CONFIG: '{{ ansible_env.ProgramData }}\Splunk\OpenTelemetry Collector\custom_config.yml'
       SPLUNK_INGEST_URL: https://fake-splunk-ingest.com
       SPLUNK_API_URL: https://fake-splunk-api.com
@@ -15,11 +15,10 @@
       SPLUNK_BALLAST_SIZE_MIB: "100"
       MY_CUSTOM_VAR1: value1
       MY_CUSTOM_VAR2: value2
-      SIGNALFX_DOTNET_TRACER_HOME: '{{ ansible_env.ProgramFiles }}\SignalFx\.NET Tracing\'
-    dotnet_reg_values:
-      COR_ENABLE_PROFILING: "true"
+    iis_reg_values:
+      COR_ENABLE_PROFILING: "1"
       COR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"
-      CORECLR_ENABLE_PROFILING: "true"
+      CORECLR_ENABLE_PROFILING: "1"
       CORECLR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"
       SIGNALFX_ENV: test-environment
       SIGNALFX_SERVICE_NAME: test-service-name
@@ -28,6 +27,8 @@
       SIGNALFX_GLOBAL_TAGS: splunk.zc.method:signalfx-dotnet-tracing-1.0.0,dotnet-tag:dotnet-tag-value
       SIGNALFX_DOTNET_VAR1: dotnet-value1
       SIGNALFX_DOTNET_VAR2: dotnet-value2
+    machine_reg_values:
+      SIGNALFX_DOTNET_TRACER_HOME: '{{ ansible_env.ProgramFiles }}\SignalFx\.NET Tracing\'
   tasks:
     - name: Check splunk-otel-collector service
       ansible.windows.win_service:
@@ -36,10 +37,6 @@
       check_mode: yes
       register: service_status
 
-    - name: Assert splunk-otel-collector service status
-      assert:
-        that: not service_status.changed
-
     - name: Check fluentdwinsvc service
       ansible.windows.win_service:
         name: fluentdwinsvc
@@ -124,12 +121,23 @@
     - name: Verify IIS env vars
       assert:
         that: (item.key + '=' + item.value) in iis_env.value
-      loop: "{{ dotnet_reg_values | dict2items }}"
+      loop: "{{ iis_reg_values | dict2items }}"
+
+    - name: Get splunk-otel-collector service env vars
+      ansible.windows.win_reg_stat:
+        path: HKLM:\SYSTEM\CurrentControlSet\Services\splunk-otel-collector
+        name: Environment
+      register: collector_env
+
+    - name: Verify splunk-otel-collector service env vars
+      assert:
+        that: (item.key + '=' + item.value) in collector_env.value
+      loop: "{{ collector_reg_values | dict2items }}"
 
     - name: Verify env vars
       include_tasks: ../shared/verify_registry_key.yml
       vars:
         path: HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
         name: "{{ item.key }}"
         value: "{{ item.value }}"
-      loop: "{{ reg_values | combine(dotnet_reg_values) | dict2items }}"
+      loop: "{{ machine_reg_values | dict2items }}"

deployments/ansible/molecule/default/windows-verify.yml

@@ -4,7 +4,7 @@
   gather_facts: true
   become: no
   vars:
-    reg_values:
+    collector_reg_values:
       SPLUNK_CONFIG: '{{ ansible_env.ProgramData }}\Splunk\OpenTelemetry Collector\agent_config.yaml'
       SPLUNK_ACCESS_TOKEN: fake-token
       SPLUNK_REALM: fake-realm
@@ -21,10 +21,6 @@
       check_mode: yes
       register: service_status
 
-    - name: Assert splunk-otel-collector service status
-      assert:
-        that: not service_status.changed
-
     - name: Check fluentdwinsvc service
       ansible.windows.win_service:
         name: fluentdwinsvc
@@ -36,10 +32,13 @@
       assert:
         that: not service_status.exists
 
-    - name: Verify env vars
-      include_tasks: ../shared/verify_registry_key.yml
-      vars:
-        path: HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
-        name: "{{ item.key }}"
-        value: "{{ item.value }}"
-      loop: "{{ reg_values | dict2items }}"
+    - name: Get splunk-otel-collector service env vars
+      ansible.windows.win_reg_stat:
+        path: HKLM:\SYSTEM\CurrentControlSet\Services\splunk-otel-collector
+        name: Environment
+      register: collector_env
+
+    - name: Verify splunk-otel-collector service env vars
+      assert:
+        that: (item.key + '=' + item.value) in collector_env.value
+      loop: "{{ collector_reg_values | dict2items }}"

deployments/ansible/molecule/with_instrumentation/windows-verify.yml

@@ -4,7 +4,7 @@
   gather_facts: true
   become: no
   vars:
-    reg_values:
+    collector_reg_values:
       SPLUNK_CONFIG: '{{ ansible_env.ProgramData }}\Splunk\OpenTelemetry Collector\agent_config.yaml'
       SPLUNK_ACCESS_TOKEN: fake-token
       SPLUNK_REALM: fake-realm
@@ -13,16 +13,17 @@
       SPLUNK_HEC_URL: https://ingest.fake-realm.signalfx.com/v1/log
       SPLUNK_INGEST_URL: https://ingest.fake-realm.signalfx.com
       SPLUNK_TRACE_URL: https://ingest.fake-realm.signalfx.com/v2/trace
-      SIGNALFX_DOTNET_TRACER_HOME: '{{ ansible_env.ProgramFiles }}\SignalFx\.NET Tracing\'
-    dotnet_reg_values:
-      COR_ENABLE_PROFILING: "true"
+    iis_reg_values:
+      COR_ENABLE_PROFILING: "1"
       COR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"
-      CORECLR_ENABLE_PROFILING: "true"
+      CORECLR_ENABLE_PROFILING: "1"
       CORECLR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"
       SIGNALFX_ENV: ""
       SIGNALFX_SERVICE_NAME: ""
       SIGNALFX_PROFILER_ENABLED: "false"
       SIGNALFX_PROFILER_MEMORY_ENABLED: "false"
+    machine_reg_values:
+      SIGNALFX_DOTNET_TRACER_HOME: '{{ ansible_env.ProgramFiles }}\SignalFx\.NET Tracing\'
   tasks:
     - name: Check splunk-otel-collector service
       ansible.windows.win_service:
@@ -31,10 +32,6 @@
       check_mode: yes
       register: service_status
 
-    - name: Assert splunk-otel-collector service status
-      assert:
-        that: not service_status.changed
-
     - name: Get installed signalfx-dotnet-tracing MSI version
       ansible.windows.win_shell: |
         $msi_version = ""
@@ -44,19 +41,11 @@
         echo $msi_version
       register: msi_version
 
-    - name: Add SIGNALFX_GLOBAL_TAGS to dotnet_reg_values
+    - name: Add SIGNALFX_GLOBAL_TAGS to iis_reg_values
       set_fact:
-        dotnet_reg_values: |-
+        iis_reg_values: |-
           {%- set tags = "splunk.zc.method:signalfx-dotnet-tracing-" + (msi_version.stdout | trim) -%}
-          {{ dotnet_reg_values | combine({"SIGNALFX_GLOBAL_TAGS": tags}) }}
-
-    - name: Verify collector env vars
-      include_tasks: ../shared/verify_registry_key.yml
-      vars:
-        path: HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
-        name: "{{ item.key }}"
-        value: "{{ item.value }}"
-      loop: "{{ reg_values | dict2items }}"
+          {{ iis_reg_values | combine({"SIGNALFX_GLOBAL_TAGS": tags}) }}
 
     - name: Get IIS env vars
       ansible.windows.win_reg_stat:
@@ -67,12 +56,31 @@
     - name: Verify IIS env vars
       assert:
         that: (item.key + '=' + item.value) in iis_env.value
-      loop: "{{ dotnet_reg_values | dict2items }}"
+      loop: "{{ iis_reg_values | dict2items }}"
+
+    - name: Get splunk-otel-collector service env vars
+      ansible.windows.win_reg_stat:
+        path: HKLM:\SYSTEM\CurrentControlSet\Services\splunk-otel-collector
+        name: Environment
+      register: collector_env
+
+    - name: Verify splunk-otel-collector service env vars
+      assert:
+        that: (item.key + '=' + item.value) in collector_env.value
+      loop: "{{ collector_reg_values | dict2items }}"
 
     - name: Verify .NET tracing env vars were not added to the system
       include_tasks: ../shared/verify_registry_key.yml
       vars:
         path: HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
         name: "{{ item.key }}"
         exists: false
-      loop: "{{ dotnet_reg_values | dict2items }}"
+      loop: "{{ iis_reg_values | dict2items }}"
+
+    - name: Verify .NET tracing MSI env vars were added to the system
+      include_tasks: ../shared/verify_registry_key.yml
+      vars:
+        path: HKLM:\SYSTEM\CurrentControlSet\Control\Session Manager\Environment
+        name: "{{ item.key }}"
+        value: "{{ item.value }}"
+      loop: "{{ machine_reg_values | dict2items }}"

deployments/ansible/roles/collector/README.md

@@ -335,9 +335,9 @@ For proxy options, see the [Windows Proxy](#windows-proxy) section.
   `signalfx_dotnet_auto_instrumentation_additional_options` option to
   enable/configure auto instrumentation for ***only*** IIS applications:
   ```yaml
-  COR_ENABLE_PROFILING: true  # Required
+  COR_ENABLE_PROFILING: "1"  # Required
   COR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"  # Required
-  CORECLR_ENABLE_PROFILING: true  # Required
+  CORECLR_ENABLE_PROFILING: "1"  # Required
   CORECLR_PROFILER: "{B4C89B0F-9908-4F73-9F59-0D77C5A06874}"  # Required
   SIGNALFX_ENV: "{{ signalfx_dotnet_auto_instrumentation_environment }}"
   SIGNALFX_GLOBAL_TAGS: "{{ signalfx_dotnet_auto_instrumentation_global_tags }}"

deployments/ansible/roles/collector/handlers/main.yml

@@ -25,9 +25,24 @@
     - (start_service | default(true) | bool)
 
 - name: Restart Splunk OpenTelemetry Collector for windows
-  ansible.windows.win_service:
-    name: splunk-otel-collector
-    state: restarted
+  ansible.windows.win_shell: |
+    Try {
+      Restart-Service splunk-otel-collector
+    } Catch {
+      # Try to get some more helpful information given that the error message is not very helpful
+      Write-Host "Error restarting splunk-otel-collector service: $_"
+
+      Write-Host "Splunk OpenTelemetry Collector service registry entry:"
+      Get-Item HKLM:\SYSTEM\CurrentControlSet\Services\splunk-otel-collector
+
+      Write-Host "Last 15 Application log events:"
+      Get-WinEvent -Log Application -MaxEvents 15 | Format-List TimeCreated, ProviderName, Message
+
+      Write-Host "Last 15 System log events:"
+      Get-WinEvent -Log System -MaxEvents 15 | Format-List TimeCreated, ProviderName, Message
+
+      Throw $_
+    }
   listen: "restart windows splunk-otel-collector"
   when:
     - (start_service | default(true) | bool)

deployments/ansible/roles/collector/tasks/otel_win_install.yml

@@ -37,7 +37,6 @@
   ansible.windows.win_package:
     path: "{{otel_msi_package.dest}}"
     state: present
-  notify: "restart windows splunk-otel-collector"
 
 - name: Merge custom config into the default config
   ansible.builtin.import_tasks: config_override.yml
@@ -48,11 +47,9 @@
     content: '{{ updated_config | to_nice_yaml (indent=2) }}'
     dest: "{{ splunk_otel_collector_config }}"
   when: splunk_config_override != ''
-  notify: "restart windows splunk-otel-collector"
 
 - name: Push Custom Config file for splunk-otel-collector, If provided
   ansible.windows.win_template:
     src: "{{splunk_otel_collector_config_source}}"
     dest: "{{splunk_otel_collector_config}}"
   when: splunk_otel_collector_config_source != ""
-  notify: "restart windows splunk-otel-collector"