[Release 1.0] Add Elemental Reset capabilities to nodes provisioned v…

…ia EIB (#463)

Adding Elemental Reset capabilities to nodes provisioned via EIB (#449)

* Added functionality to support Elemental node reset

* fixing eof issues

* fixing permissions

* fixing test

* updating release notes

* Moving to elemental_node_cleanup.sh

Co-authored-by: Rhys Oxenham <[email protected]>

RELEASE_NOTES.md

@@ -5,6 +5,7 @@
 ## General
 
 * Added the ability to consume both 512/4096 byte sector size disk input base-images
+* Added the ability to leverage Elemental node reset for unmanaged operating systems
 * Added version command and version marker on CRB images
 
 ## API

pkg/combustion/elemental_test.go

@@ -63,4 +63,8 @@ func TestWriteElementalCombustionScript(t *testing.T) {
 	require.NoError(t, err)
 	found := string(foundBytes)
 	assert.Contains(t, found, "/usr/sbin/elemental-register --debug --config-path /etc/elemental/config.yaml --state-path /etc/elemental/state.yaml --install --no-toolkit")
+	assert.Contains(t, found, "/etc/systemd/system/elemental-reset.path")
+	assert.Contains(t, found, "/etc/systemd/system/elemental-reset.service")
+	assert.Contains(t, found, "mkdir -p /opt/edge/")
+	assert.Contains(t, found, "cat <<- \\EOF > /opt/edge/elemental_node_cleanup.sh")
 }

pkg/combustion/templates/31-elemental-register.sh.tpl

@@ -47,5 +47,113 @@ Environment="CATTLE_AGENT_CONFIG=/etc/rancher/elemental/agent/config.yaml"
 ExecStart=/usr/sbin/elemental-system-agent sentinel
 EOF
 
+cat <<- EOF > /etc/systemd/system/elemental-reset.path
+[Path]
+PathModified=/var/lib/elemental/.unmanaged_reset
+
+[Install]
+WantedBy=multi-user.target
+EOF
+
+cat <<- EOF > /etc/systemd/system/elemental-reset.service
+[Unit]
+Description=Elemental Reset for Unmanaged Hosts
+Wants=network-online.target
+After=network-online.target
+
+[Service]
+Type=simple
+ExecStart=/opt/edge/elemental_node_cleanup.sh -u
+ExecStartPost=/usr/bin/rm -f /var/lib/elemental/.unmanaged_reset
+EOF
+
+systemctl enable elemental-reset.path || true
 systemctl enable elemental-register-systemd.service || true
 systemctl enable elemental-system-agent.service || true
+
+mkdir -p /opt/edge/
+cat <<- \EOF > /opt/edge/elemental_node_cleanup.sh
+#!/usr/bin/env bash
+# SUSE Edge Elemental Node Reset Script
+# Copyright 2024 SUSE Software Solutions
+
+# This script attempts to cleanup a node that has been deployed via Edge Image
+# Builder with the integrations for Elemental registration; in other words,
+# vanilla SLE Micro 5.5, *not* SLE Micro for Rancher (also known as Elemental
+# Teal), that has used the "--no-toolkit" registration option.
+#
+# The default behaviour in Rancher/Elemental is that in the event that a
+# cluster is deleted in Rancher, the Kubernetes cluster running on a node (or
+# set of nodes) will not be automatically cleaned up; the cluster will be
+# orphaned and will remain running. Furthermore, the Elemental MachineInventory
+# will be removed, so it's no longer visible in the list of registered nodes.
+#
+# This script cleans up the installed Kubernetes cluster so no traces remain
+# and forces a re-registration with the original Elemental registration config.
+#
+# WARNING: This script *will* cause data loss as it removes all Kubernetes
+#          persistent data. There is also an unattended switch for automated
+#          reset. You have been warned!
+
+UNATTENDED=false
+
+while getopts 'u' OPTION; do
+    case "${OPTION}" in
+        u)
+            UNATTENDED=true
+            ;;
+    esac
+done
+
+if [ $UNATTENDED = "false" ] ;
+then
+    echo "============================================"
+    echo "SUSE Edge Node Cleanup for Elemental Systems"
+    echo -e "============================================\n"
+    echo -n "WARNING: This script will remove all Kubernetes files and will"
+    echo -e " cause data loss!\n"
+    while true; do
+            read -p "Are you sure you wish to proceed [y/N]? " yn
+            case $yn in
+                [Yy] ) break;;
+                [Nn] ) exit;;
+                * ) exit 0;;
+            esac
+        done
+fi
+
+# If we reach this point, we're deleting data and re-registering.
+
+# Stop both the elemental and rancher-system-agents via systemd
+systemctl kill --signal=SIGKILL elemental-system-agent
+systemctl kill --signal=SIGKILL rancher-system-agent
+
+# Kill and uninstall all rke2 services
+if [ -x /opt/rke2/bin/rke2-uninstall.sh ];
+then
+    /opt/rke2/bin/rke2-killall.sh
+    /opt/rke2/bin/rke2-uninstall.sh
+fi
+
+# Kill and uninstall all k3s services
+if command -v k3s-killall.sh &> /dev/null; then k3s-killall.sh; fi
+if command -v k3s-uninstall.sh &> /dev/null; then k3s-uninstall.sh; fi
+
+# Remove the rancher-system-agent as this gets reinstalled via Elemental
+if [ -x /opt/rancher-system-agent/bin/rancher-system-agent-uninstall.sh ];
+then
+    sh /opt/rancher-system-agent/bin/rancher-system-agent-uninstall.sh
+    rm -rf /opt/rancher-system-agent
+fi
+
+# Clean up all old configuration directories and Elemental state
+rm -rf /etc/rancher
+rm -rf /var/lib/rancher
+rm -f /etc/elemental/state.yaml
+
+# Re-register the node via Elemental using the original Elemental config
+# by restarting the Elemental registration service via systemd
+systemctl restart elemental-register-systemd.service
+EOF
+
+chmod a+x /opt/edge/elemental_node_cleanup.sh