clusterconf/monitor: relax vps free space rules except on node23

We'll rely on datasets being expanded by vpsAdmin. The alerts should
fire only should that fail. node23 is set aside until it is updated.

modules/clusterconf/monitor/rules/nodes.nix

@@ -686,7 +686,7 @@
       {
         alert = "VpsOsCritFreeSpace";
         expr = ''osctl_container_dataset_avail_bytes{job="nodes"} < 256 * 1024 * 1024'';
-        for = "2m";
+        for = "5m";
         labels = {
           alertclass = "vpsdiskspace";
           severity = "critical";
@@ -706,6 +706,26 @@
       {
         alert = "VpsOsFatalFreeSpace";
         expr = ''osctl_container_dataset_avail_bytes{alias=~"node21.prg|node22.prg|node23.prg|node24.prg"} < 256 * 1024 * 1024'';
+        for = "5m";
+        labels = {
+          alertclass = "vpsdiskspace";
+          severity = "fatal";
+          frequency = "2m";
+        };
+        annotations = {
+          summary = "Not enough free space (instance {{ $labels.instance }})";
+          description = ''
+            VPS has less than 256 MB of diskspace left
+
+            VALUE = {{ $value }}
+            LABELS: {{ $labels }}
+          '';
+        };
+      }
+
+      {
+        alert = "VpsOsFatalFreeSpaceNode23";
+        expr = ''osctl_container_dataset_avail_bytes{alias="node23.prg"} < 256 * 1024 * 1024'';
         for = "2m";
         labels = {
           alertclass = "vpsdiskspace";