Implement basic metrics

README.md

@@ -17,7 +17,7 @@ Features
 - [ ] Volume modes
     - [x] `Filesystem` mode
     - [ ] `Block` mode
-- [ ] Volume metrics
+- [x] Volume metrics
 - [ ] Supports fsTypes
 - [ ] Online expansion: If fs supports it (e.g. ext4, btrfs)
 - [ ] Online shrinking: If fs supports it (e.g. btrfs)

metrics.py

@@ -0,0 +1,61 @@
+import json
+import os
+import threading
+from os.path import basename
+
+from prometheus_client import Gauge
+from prometheus_client.exposition import start_http_server
+
+import rawfile_util
+from rawfile_util import attached_loops
+from util import run_out
+
+VOLUME_ID = "volume_id"
+
+fs_size = Gauge(
+    "rawfile_filesystem_size_bytes", "Filesystem size in bytes.", [VOLUME_ID]
+)
+fs_free = Gauge(
+    "rawfile_filesystem_avail_bytes", "Filesystem free space in bytes", [VOLUME_ID]
+)
+dev_size = Gauge("rawfile_device_size_bytes", "Device size in bytes.", [VOLUME_ID])
+dev_free = Gauge(
+    "rawfile_device_free_bytes", "Device free space in bytes.", [VOLUME_ID]
+)
+
+
+def collect_stats():
+    blockdevices = json.loads(run_out("lsblk --json").stdout.decode())["blockdevices"]
+
+    def dev_to_mountpoint(dev_name):
+        dev_name = basename(dev_name)
+        matches = list(filter(lambda bd: bd["name"] == dev_name, blockdevices))
+        if len(matches) == 0:
+            return None
+        return matches[0]["mountpoint"]
+
+    for volume_id in rawfile_util.list_all_volumes():
+        img_file = rawfile_util.img_file(volume_id)
+        labels = {VOLUME_ID: volume_id}
+        dev_stat = img_file.stat()
+        dev_size.labels(**labels).set(dev_stat.st_size)
+        dev_free.labels(**labels).set(
+            dev_stat.st_size - dev_stat.st_blocks * dev_stat.st_blksize
+        )
+        for dev in attached_loops(img_file):
+            mountpoint = dev_to_mountpoint(dev)
+            if mountpoint is None:
+                continue
+            fs_stat = os.statvfs(mountpoint)
+            fs_size.labels(**labels).set(fs_stat.f_frsize * fs_stat.f_blocks)
+            fs_free.labels(**labels).set(fs_stat.f_frsize * fs_stat.f_bfree)
+            break
+
+
+def expose_metrics():
+    def collector_loop():
+        collect_stats()
+        threading.Timer(10, collector_loop).start()
+
+    collector_loop()
+    start_http_server(9100)

rawfile.py

@@ -7,6 +7,7 @@ import grpc
 
 import rawfile_servicer
 from csi import csi_pb2_grpc
+from metrics import expose_metrics
 
 
 @click.group()
@@ -17,7 +18,10 @@ def cli():
 @cli.command()
 @click.option("--endpoint", envvar="CSI_ENDPOINT", default="0.0.0.0:5000")
 @click.option("--nodeid", envvar="NODE_ID")
-def csi_driver(endpoint, nodeid):
+@click.option("--enable-metrics/--disable-metrics", default=True)
+def csi_driver(endpoint, nodeid, enable_metrics):
+    if enable_metrics:
+        expose_metrics()
     server = grpc.server(futures.ThreadPoolExecutor(max_workers=10))
     csi_pb2_grpc.add_IdentityServicer_to_server(
         rawfile_servicer.RawFileIdentityServicer(), server

rawfile_util.py

@@ -1,4 +1,6 @@
+import glob
 import json
+from os.path import basename, dirname
 from pathlib import Path
 
 from consts import DATA_DIR
@@ -53,3 +55,8 @@ def attach_loop(file) -> str:
             return devs[0]
         next_loop()
         run(f"losetup --direct-io=on -f {file}")
+
+
+def list_all_volumes():
+    metas = glob.glob(f"{DATA_DIR}/*/disk.meta")
+    return [basename(dirname(meta)) for meta in metas]

requirements.in

@@ -4,3 +4,4 @@ click
 pyyaml
 pykube-ng
 munch
+prometheus_client

requirements.txt

@@ -11,6 +11,7 @@ grpcio-tools==1.28.1      # via -r requirements.in
 grpcio==1.28.1            # via -r requirements.in, grpcio-tools
 idna==2.9                 # via requests
 munch==2.5.0              # via -r requirements.in
+prometheus-client==0.7.1  # via -r requirements.in
 protobuf==3.11.3          # via grpcio-tools
 pykube-ng==20.4.1         # via -r requirements.in
 pyyaml==5.3.1             # via -r requirements.in, pykube-ng