fix(metrics): improve cleanup of old service

The core-update action was not correctly shutdown old node-exporter instances and node-monitor services.

Author: gsanchietti

core/imageroot/etc/systemd/system/node_exporter.service

@@ -7,6 +7,10 @@ Environment=PODMAN_SYSTEMD_UNIT=%n
 EnvironmentFile=-/etc/nethserver/core.env
 WorkingDirectory=/var/lib/nethserver/node/state
 Restart=always
+RestartSec=3
+TimeoutStartSec=120
+# node_exporter exits with 143 on SIGTERM:
+SuccessExitStatus=143
 ExecStartPre=/bin/rm -f %t/%N.pid %t/%N.cid
 ExecStartPre=/usr/bin/mkdir -p /run/node_exporter
 ExecStart=/usr/bin/podman run \

core/imageroot/var/lib/nethserver/node/update-core.d/50start_node_exporter

@@ -11,5 +11,7 @@ set -e
 if systemctl is-enabled --quiet node_exporter ; then
     systemctl restart node_exporter.service || :
 else
+    # Make sure to stop all node_exporterX instances
+    systemctl stop node_exporter*.service || :
     systemctl enable --now node_exporter.service
 fi

core/imageroot/var/lib/nethserver/node/update-core.d/80node_monitor

@@ -6,11 +6,12 @@
 #
 
 exec 1>&2
-set -e
 
 if systemctl -q is-active node-monitor.service; then
     echo "Disabling node-monitor as service"
+    systemctl stop node-monitor
+    # The disable may fail if the update-core has already removed the service file
     systemctl disable --now node-monitor
-    rm -f /etc/systemd/system/node-monitor.service
+    rm -fv /etc/systemd/system/node-monitor.service
     systemctl daemon-reload
 fi