diff --git a/ops/monitoring/doc/index.md b/ops/monitoring/doc/index.md
new file mode 100644
index 00000000..b17287cd
--- /dev/null
+++ b/ops/monitoring/doc/index.md
@@ -0,0 +1,38 @@
+Monitoring
+==========
+
+Setting up monitoring in hscloud is a work in progress.
+
+Components
+----------
+
+Currently we have a per-cluster setup with prometheus scraping Kubernetes nodes
+(kubeletes) for kubelet metrics and cAdvisor metrics.
+
+    .-----------------------------------------------------------.
+    |                        k0.hswaw.net                       |
+    |-----------------------------------------------------------|
+    |  .---------------------.                                  |
+    |  | ns: metrics-cluster |    .--------------------------.  |
+    |  |---------------------|    |  kubernetes.svc.cluster  |  |
+    |  | prometheus          |--> | apiserver proxy to nodes |  |
+    |  '---------------------'    '--------------------------'  |
+    |                                         |                 |
+    '---------------------------------------- v ----------------'
+                                  .---------------------.
+                                  |   bc0n01.hswaw.net  |-.
+                      Kubernetes  |---------------------| |-.
+                           Nodes  |  /metrics           |-| |
+                                  |  /metrics/cadvisor  | |-|
+                                  '---------------------' | |
+                                    '---------------------' |
+                                      '---------------------'
+
+Everything else (dashboard, aggregation, user metrics) is a work in progress.
+
+Legacy
+------
+
+There is a legacy prometheus/grafana VM on https://metrics.hackerspace.pl/. The
+certificate is expired, but it Generally Works, and will be kept going until
+its functionality is migrated to hscloud.
diff --git a/ops/monitoring/k0.jsonnet b/ops/monitoring/k0.jsonnet
new file mode 100644
index 00000000..028a4630
--- /dev/null
+++ b/ops/monitoring/k0.jsonnet
@@ -0,0 +1,11 @@
+local lib = import "lib.libsonnet";
+
+{
+    cluster: lib.Cluster("k0") {
+        cfg+: {
+            storageClasses+: {
+                prometheus: "waw-hdd-redundant-3",
+            },
+        },
+    },
+}
diff --git a/ops/monitoring/lib.libsonnet b/ops/monitoring/lib.libsonnet
new file mode 100644
index 00000000..61f49b49
--- /dev/null
+++ b/ops/monitoring/lib.libsonnet
@@ -0,0 +1,5 @@
+local cluster = import "lib/cluster.libsonnet";
+
+{
+    Cluster: cluster.Cluster,
+}
diff --git a/ops/monitoring/kube.jsonnet b/ops/monitoring/lib/cluster.libsonnet
similarity index 96%
rename from ops/monitoring/kube.jsonnet
rename to ops/monitoring/lib/cluster.libsonnet
index 919ca7d7..9b64f05c 100644
--- a/ops/monitoring/kube.jsonnet
+++ b/ops/monitoring/lib/cluster.libsonnet
@@ -1,8 +1,9 @@
-local kube = import "../../kube/kube.libsonnet";
+local kube = import "../../../kube/kube.libsonnet";
 
 {
-    local top = self,
-    // Per-cluster components
+    // Cluster sets up all cluster-specific monitoring resources in their own namespace.
+    // Currently this consists of a prometheus server that scrapes k8s nodes for kubelet
+    // and cAdvisor metrics.
     Cluster(name):: {
         local cluster = self,
         local cfg = cluster.cfg,
@@ -15,7 +16,7 @@ local kube = import "../../kube/kube.libsonnet";
             },
 
             storageClasses: {
-                prometheus: "waw-hdd-redundant-3",
+                prometheus: error "storageClasses.prometheus must be set",
             },
         },
 
@@ -223,6 +224,4 @@ local kube = import "../../kube/kube.libsonnet";
             },
         },
     },
-
-    k0: top.Cluster("k0"),
 }