Enable "Storage Capacity Tracking"

.ci/e2e-test/rawfile-driver.yaml

@@ -21,4 +21,7 @@ DriverInfo:
     nodeExpansion: true
     onlineExpansion: true
     volumeLimits: false
+    singleNodeVolume: true
+    topology: true
+    capacity: true
 InlineVolumes: []

.ci/e2e-test/setup.sh

@@ -2,8 +2,8 @@
 set -ex
 source .ci/common
 
-K8S_VERSION=1.19.12
-MINIKUBE_VERSION=1.19.0
+K8S_VERSION=1.21.5
+MINIKUBE_VERSION=1.21.0
 curl -Lo kubectl https://storage.googleapis.com/kubernetes-release/release/v${K8S_VERSION}/bin/linux/amd64/kubectl && chmod +x kubectl && sudo mv kubectl /usr/local/bin/
 sudo apt update && sudo apt install -y conntrack
 curl -Lo minikube https://storage.googleapis.com/minikube/releases/v${MINIKUBE_VERSION}/minikube-linux-amd64 && chmod +x minikube && sudo mv minikube /usr/local/bin/

README.md

@@ -7,7 +7,7 @@ Kubernetes LocalPVs on Steroids
 
 Prerequisite
 ---
-- Kubernetes: 1.19+
+- Kubernetes: 1.21+
 
 Install
 ---

bd2fs.py

@@ -211,6 +211,10 @@ class Bd2FsControllerServicer(csi_pb2_grpc.ControllerServicer):
     def DeleteVolume(self, request, context):
         return self.bds.DeleteVolume(request, context)
 
+    @log_grpc_request
+    def GetCapacity(self, request, context):
+        return self.bds.GetCapacity(request, context)
+
     @log_grpc_request
     def ControllerExpandVolume(self, request, context):
         response = self.bds.ControllerExpandVolume(request, context)

deploy/charts/rawfile-csi/templates/00-driver.yaml

@@ -6,5 +6,6 @@ spec:
   attachRequired: false
   podInfoOnMount: true
   fsGroupPolicy: File
+  storageCapacity: true
   volumeLifecycleModes:
     - Persistent

deploy/charts/rawfile-csi/templates/00-rbac.yaml

@@ -40,6 +40,15 @@ rules:
   - apiGroups: ["storage.k8s.io"]
     resources: ["volumeattachments"]
     verbs: ["get", "list", "watch"]
+  - apiGroups: ["storage.k8s.io"]
+    resources: ["csistoragecapacities"]
+    verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
+  - apiGroups: [""]
+    resources: ["pods"]
+    verbs: ["get"]
+  - apiGroups: ["apps"]
+    resources: ["daemonsets"]
+    verbs: ["get"]
 ---
 kind: ClusterRoleBinding
 apiVersion: rbac.authorization.k8s.io/v1

deploy/charts/rawfile-csi/templates/01-controller-plugin.yaml

@@ -63,20 +63,6 @@ spec:
               containerPort: 9808
           resources:
             {{- toYaml .Values.controller.resources | nindent 12 }}
-        - name: external-provisioner
-          image: k8s.gcr.io/sig-storage/csi-provisioner:v2.2.2
-          imagePullPolicy: IfNotPresent
-          args:
-            - "--csi-address=$(ADDRESS)"
-            - "--feature-gates=Topology=true"
-            - "--strict-topology"
-            - "--timeout=120s"
-          env:
-            - name: ADDRESS
-              value: /csi/csi.sock
-          volumeMounts:
-            - name: socket-dir
-              mountPath: /csi
         - name: external-resizer
           image: k8s.gcr.io/sig-storage/csi-resizer:v1.2.0
           imagePullPolicy: IfNotPresent

deploy/charts/rawfile-csi/templates/01-node-plugin.yaml

@@ -123,3 +123,33 @@ spec:
             requests:
               cpu: 10m
               memory: 100Mi
+        - name: external-provisioner
+          image: k8s.gcr.io/sig-storage/csi-provisioner:v2.2.2
+          imagePullPolicy: IfNotPresent
+          args:
+            - "--csi-address=$(ADDRESS)"
+            - "--feature-gates=Topology=true"
+            - "--strict-topology"
+            - "--immediate-topology=false"
+            - "--timeout=120s"
+            - "--enable-capacity=true"
+            - "--capacity-ownerref-level=1" # DaemonSet
+            - "--node-deployment=true"
+          env:
+            - name: ADDRESS
+              value: /csi/csi.sock
+            - name: NODE_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: spec.nodeName
+            - name: NAMESPACE
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.namespace
+            - name: POD_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: metadata.name
+          volumeMounts:
+            - name: socket-dir
+              mountPath: /csi

rawfile_servicer.py

@@ -11,7 +11,7 @@ from declarative import be_symlink, be_absent
 from fs_util import device_stats, mountpoint_to_dev
 from orchestrator.k8s import volume_to_node, run_on_node
 from rawfile_util import attach_loop, detach_loops
-from remote import init_rawfile, scrub, expand_rawfile
+from remote import init_rawfile, scrub, get_capacity, expand_rawfile
 from util import log_grpc_request, run
 
 NODE_NAME_TOPOLOGY_KEY = "hostname"
@@ -134,6 +134,7 @@ class RawFileControllerServicer(csi_pb2_grpc.ControllerServicer):
         return csi_pb2.ControllerGetCapabilitiesResponse(
             capabilities=[
                 Cap(rpc=Cap.RPC(type=Cap.RPC.CREATE_DELETE_VOLUME)),
+                Cap(rpc=Cap.RPC(type=Cap.RPC.GET_CAPACITY)),
                 Cap(rpc=Cap.RPC(type=Cap.RPC.EXPAND_VOLUME)),
             ]
         )
@@ -186,10 +187,7 @@ class RawFileControllerServicer(csi_pb2_grpc.ControllerServicer):
             )
 
         try:
-            run_on_node(
-                init_rawfile.as_cmd(volume_id=request.name, size=size),
-                node=node_name,
-            )
+            init_rawfile(volume_id=request.name, size=size),
         except CalledProcessError as exc:
             if exc.returncode == RESOURCE_EXHAUSTED_EXIT_CODE:
                 context.abort(
@@ -210,10 +208,15 @@ class RawFileControllerServicer(csi_pb2_grpc.ControllerServicer):
 
     @log_grpc_request
     def DeleteVolume(self, request, context):
-        node_name = volume_to_node(request.volume_id)
-        run_on_node(scrub.as_cmd(volume_id=request.volume_id), node=node_name)
+        scrub(volume_id=request.volume_id)
         return csi_pb2.DeleteVolumeResponse()
 
+    @log_grpc_request
+    def GetCapacity(self, request, context):
+        return csi_pb2.GetCapacityResponse(
+            available_capacity=get_capacity(),
+        )
+
     @log_grpc_request
     def ControllerExpandVolume(self, request, context):
         volume_id = request.volume_id

remote.py

@@ -1,7 +1,6 @@
 from util import remote_fn
 
 
-@remote_fn
 def scrub(volume_id):
     import time
     import rawfile_util
@@ -17,18 +16,18 @@ def scrub(volume_id):
     rawfile_util.gc_if_needed(volume_id, dry_run=False)
 
 
-@remote_fn
 def init_rawfile(volume_id, size):
     import time
-    import rawfile_util
-    from volume_schema import LATEST_SCHEMA_VERSION
+    from subprocess import CalledProcessError
     from pathlib import Path
 
+    import rawfile_util
+    from volume_schema import LATEST_SCHEMA_VERSION
     from util import run
     from consts import RESOURCE_EXHAUSTED_EXIT_CODE
 
     if rawfile_util.get_capacity() < size:
-        exit(RESOURCE_EXHAUSTED_EXIT_CODE)
+        raise CalledProcessError(returncode=RESOURCE_EXHAUSTED_EXIT_CODE, cmd="")
 
     img_dir = rawfile_util.img_dir(volume_id)
     img_dir.mkdir(exist_ok=True)
@@ -48,6 +47,12 @@ def init_rawfile(volume_id, size):
     run(f"truncate -s {size} {img_file}")
 
 
+def get_capacity():
+    import rawfile_util
+
+    return rawfile_util.get_capacity()
+
+
 @remote_fn
 def expand_rawfile(volume_id, size):
     import rawfile_util