refactor(linux): move cgroup telemetry service enable and start to nodePrep

parts/linux/cloud-init/artifacts/cse_config.sh

@@ -823,3 +823,20 @@ get_compute_sku() {
     fi
     echo "$vm_sku"
 }
+
+enableAndStartCgroupTelemetry() {
+    # Don't enable cgroup-memory-telemetry on AzureLinux due to incompatibility with cgroup2fs driver and absence of required azure.slice directory
+    if ! isMarinerOrAzureLinux "$OS"; then
+        systemctlEnableAndStartNoBlock cgroup-memory-telemetry.timer 30 || exit $ERR_START_CGROUP_TELEMETRY
+        systemctlEnableAndStartNoBlock cgroup-memory-telemetry.service 30 || exit $ERR_START_CGROUP_TELEMETRY
+    fi
+
+    CGROUP_VERSION=$(stat -fc %T /sys/fs/cgroup)
+    if [ "$CGROUP_VERSION" = "cgroup2fs" ]; then
+        systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.timer 30 || exit $ERR_START_CGROUP_TELEMETRY
+        systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.service 30 || exit $ERR_START_CGROUP_TELEMETRY
+
+        systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.timer 30 || exit $ERR_START_CGROUP_TELEMETRY
+        systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.service 30 || exit $ERR_START_CGROUP_TELEMETRY
+    fi
+}

parts/linux/cloud-init/artifacts/cse_helpers.sh

@@ -170,6 +170,8 @@ ERR_CHRONY_NTP_SYNC_FAIL=245 # Chrony could not synchronize with the configured
 ERR_CHRONY_CONFIG_FAIL=246 # Chrony could not be configured for Intel TDX
 # -----------------------------------------------------------------------------
 
+ERR_START_CGROUP_TELEMETRY=247 # Error enabling and starting cgroup telemetry services
+
 # This probably wasn't launched via a login shell, so ensure the PATH is correct.
 [ -f /etc/profile.d/path.sh ] && . /etc/profile.d/path.sh
 

parts/linux/cloud-init/artifacts/cse_main.sh

@@ -663,6 +663,8 @@ function nodePrep {
         fi
     fi
 
+    logs_to_events "AKS.CSE.enableAndStartCgroupTelemetry" enableAndStartCgroupTelemetry
+
     if $REBOOTREQUIRED; then
         echo 'reboot required, rebooting node in 1 minute'
         /bin/bash -c "shutdown -r 1 &"

spec/parts/linux/cloud-init/artifacts/cse_config_spec.sh

@@ -12,6 +12,109 @@ Describe 'cse_config.sh'
     CSE_CONFIG_ADDONS_FILEPATH="./parts/linux/cloud-init/artifacts/cse_config_addons.sh"
     Include "./parts/linux/cloud-init/artifacts/cse_config.sh"
     Include "./parts/linux/cloud-init/artifacts/cse_helpers.sh"
+    Describe 'enableAndStartCgroupTelemetry'
+        setup_cgroup_telemetry() {
+            OS="UBUNTU"
+            OS_VARIANT=""
+            MOCK_CGROUP_VERSION="cgroup2fs"
+            FAILED_TELEMETRY_COMMAND=""
+            memory_commands="systemctlEnableAndStartNoBlock cgroup-memory-telemetry.timer 30
+systemctlEnableAndStartNoBlock cgroup-memory-telemetry.service 30"
+            cpu_pressure_commands="systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.timer 30
+systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.service 30
+systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.timer 30
+systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.service 30"
+        }
+
+        BeforeEach 'setup_cgroup_telemetry'
+
+        stat() {
+            if [ "$*" != "-fc %T /sys/fs/cgroup" ]; then
+                echo "Unexpected stat arguments: $*" >&2
+                return 1
+            fi
+            echo "$MOCK_CGROUP_VERSION"
+        }
+
+        systemctlEnableAndStartNoBlock() {
+            echo "systemctlEnableAndStartNoBlock $*"
+            [ "systemctlEnableAndStartNoBlock $*" != "$FAILED_TELEMETRY_COMMAND" ]
+        }
+
+        Describe 'OS and cgroup eligibility'
+            Parameters
+                UBUNTU "" true
+                MARINER "" false
+                MARINERKATA "" false
+                AZURELINUX "" false
+                AZURELINUXKATA "" false
+                AZURELINUX OSGUARD false
+                AZURECONTAINERLINUX "" true
+                AZURELINUX AZURECONTAINERLINUX true
+            End
+
+            It "configures eligible telemetry on cgroup v2 for $1 variant '$2'"
+                OS="$1"
+                OS_VARIANT="$2"
+                expected_commands="$cpu_pressure_commands"
+                if [ "$3" = "true" ]; then
+                    expected_commands="$memory_commands
+$cpu_pressure_commands"
+                fi
+
+                When run enableAndStartCgroupTelemetry
+
+                The status should be success
+                The output should equal "$expected_commands"
+            End
+
+            It "configures only eligible memory telemetry on cgroup v1 for $1 variant '$2'"
+                OS="$1"
+                OS_VARIANT="$2"
+                MOCK_CGROUP_VERSION="tmpfs"
+                expected_commands=""
+                if [ "$3" = "true" ]; then
+                    expected_commands="$memory_commands"
+                fi
+
+                When run enableAndStartCgroupTelemetry
+
+                The status should be success
+                The output should equal "$expected_commands"
+            End
+        End
+
+        Describe 'startup failures'
+            Parameters
+                "systemctlEnableAndStartNoBlock cgroup-memory-telemetry.timer 30" 1
+                "systemctlEnableAndStartNoBlock cgroup-memory-telemetry.service 30" 2
+                "systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.timer 30" 3
+                "systemctlEnableAndStartNoBlock cgroup-cpu-telemetry.service 30" 4
+                "systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.timer 30" 5
+                "systemctlEnableAndStartNoBlock cgroup-pressure-telemetry.service 30" 6
+            End
+
+            It "exits with the telemetry error and stops setup when '$1' fails"
+                FAILED_TELEMETRY_COMMAND="$1"
+
+                When run enableAndStartCgroupTelemetry
+
+                The status should equal "$ERR_START_CGROUP_TELEMETRY"
+                The lines of output should equal "$2"
+                The line "$2" of output should equal "$1"
+            End
+        End
+
+        It 'exits with the telemetry error when memory service startup fails on cgroup v1'
+            MOCK_CGROUP_VERSION="tmpfs"
+            FAILED_TELEMETRY_COMMAND="systemctlEnableAndStartNoBlock cgroup-memory-telemetry.service 30"
+
+            When run enableAndStartCgroupTelemetry
+
+            The status should equal "$ERR_START_CGROUP_TELEMETRY"
+            The output should equal "$memory_commands"
+        End
+    End
     Describe 'configureTransparentHugePageSystemdService'
         setup_thp_service() {
             THP_ENABLED="never"

vhdbuilder/packer/cleanup-vhd.sh

@@ -124,3 +124,5 @@ rm -f /opt/azure/disk-usage.txt
 rm -f /opt/azure/containers/image-fetcher
 # Cleanup IMDS instance metadata cache file
 rm -f /opt/azure/containers/imds_instance_metadata_cache.json
+# Cleanup Microsoft.Azure.Extensions.CustomScript to prevent unnecessary logs/telemetry payloads being persisted on the OS disk
+rm -rf /var/log/azure/Microsoft.Azure.Extensions.CustomScript

vhdbuilder/packer/install-dependencies.sh

@@ -1435,30 +1435,6 @@ fi
 configureServiceExecutionTelemetryForSystemdVersion "$(systemctl --version | awk 'NR == 1 { print $2 }')" || exit $ERR_PACKER_COPY_FILE
 systemctl daemon-reload
 
-# Disable cgroup-memory-telemetry on AzureLinux due to incompatibility with cgroup2fs driver and absence of required azure.slice directory
-if ! isMarinerOrAzureLinux "$OS"; then
-  systemctlEnableAndStart cgroup-memory-telemetry.timer 30 || exit 1
-  systemctl enable cgroup-memory-telemetry.service || exit 1
-  systemctl restart cgroup-memory-telemetry.service
-fi
-
-CGROUP_VERSION=$(stat -fc %T /sys/fs/cgroup)
-if [ "$CGROUP_VERSION" = "cgroup2fs" ]; then
-  systemctlEnableAndStart cgroup-cpu-telemetry.timer 30 || exit 1
-  systemctl enable cgroup-cpu-telemetry.service || exit 1
-  systemctl restart cgroup-cpu-telemetry.service
-
-  systemctlEnableAndStart cgroup-pressure-telemetry.timer 30 || exit 1
-  systemctl enable cgroup-pressure-telemetry.service || exit 1
-  systemctl restart cgroup-pressure-telemetry.service
-fi
-
-if [ -d "/var/log/azure/Microsoft.Azure.Extensions.CustomScript/events/" ] && [ "$(ls -A /var/log/azure/Microsoft.Azure.Extensions.CustomScript/events/)" ]; then
-  cat /var/log/azure/Microsoft.Azure.Extensions.CustomScript/events/*
-  rm -r /var/log/azure/Microsoft.Azure.Extensions.CustomScript || exit 1
-fi
-capture_benchmark "${SCRIPT_NAME}_configure_telemetry"
-
 if [ "$OS" = "$UBUNTU_OS_NAME" ]; then
   # remove snapd, which is not used by container stack
   apt_get_purge 20 30 120 snapd || exit 1