Merge pull request #825 from kasbert/master

Add disk and memory percent_used
kubernetes · Dec 15, 2023 · bdaa44e · bdaa44e
2 parents 9f639dd + 0790063
commit bdaa44e
Show file tree

Hide file tree

Showing 9 changed files with 55 additions and 0 deletions.
diff --git a/config/system-stats-monitor.json b/config/system-stats-monitor.json
@@ -44,6 +44,9 @@
       "disk/bytes_used": {
         "displayName": "disk/bytes_used"
       },
+      "disk/percent_used": {
+        "displayName": "disk/percent_used"
+      },
       "disk/io_time": {
         "displayName": "disk/io_time"
       },
@@ -88,6 +91,9 @@
       },
       "memory/unevictable_used": {
         "displayName": "memory/unevictable_used"
+      },
+      "memory/percent_used": {
+        "displayName": "memory/percent_used"
       }
     }
   },

diff --git a/config/windows-system-stats-monitor.json b/config/windows-system-stats-monitor.json
@@ -44,6 +44,9 @@
       "disk/bytes_used": {
         "displayName": "disk/bytes_used"
       },
+      "disk/percent_used": {
+        "displayName": "disk/percent_used"
+      },
       "disk/io_time": {
         "displayName": "disk/io_time"
       },
@@ -88,6 +91,9 @@
       },
       "memory/unevictable_used": {
         "displayName": "memory/unevictable_used"
+      },
+      "memory/percent_used": {
+        "displayName": "memory/percent_used"
       }
     }
   }

diff --git a/pkg/exporters/stackdriver/stackdriver_exporter.go b/pkg/exporters/stackdriver/stackdriver_exporter.go
@@ -54,6 +54,7 @@ var NPDMetricToSDMetric = map[metrics.MetricID]string{
 	metrics.CPULoad15m:              "compute.googleapis.com/guest/cpu/load_15m",
 	metrics.DiskAvgQueueLenID:       "compute.googleapis.com/guest/disk/queue_length",
 	metrics.DiskBytesUsedID:         "compute.googleapis.com/guest/disk/bytes_used",
+	metrics.DiskPercentUsedID:       "custom.googleapis.com/guest/disk/percent_used",
 	metrics.DiskIOTimeID:            "compute.googleapis.com/guest/disk/io_time",
 	metrics.DiskMergedOpsCountID:    "compute.googleapis.com/guest/disk/merged_operation_count",
 	metrics.DiskOpsBytesID:          "compute.googleapis.com/guest/disk/operation_bytes_count",
@@ -66,6 +67,7 @@ var NPDMetricToSDMetric = map[metrics.MetricID]string{
 	metrics.MemoryDirtyUsedID:       "compute.googleapis.com/guest/memory/dirty_used",
 	metrics.MemoryPageCacheUsedID:   "compute.googleapis.com/guest/memory/page_cache_used",
 	metrics.MemoryUnevictableUsedID: "compute.googleapis.com/guest/memory/unevictable_used",
+	metrics.MemoryPercentUsedID:     "custom.googleapis.com/guest/memory/percent_used",
 	metrics.ProblemCounterID:        "compute.googleapis.com/guest/system/problem_count",
 	metrics.ProblemGaugeID:          "compute.googleapis.com/guest/system/problem_state",
 	metrics.OSFeatureID:             "compute.googleapis.com/guest/system/os_feature_enabled",

diff --git a/pkg/systemstatsmonitor/disk_collector.go b/pkg/systemstatsmonitor/disk_collector.go
@@ -38,6 +38,7 @@ type diskCollector struct {
 	mOpsBytes       *metrics.Int64Metric
 	mOpsTime        *metrics.Int64Metric
 	mBytesUsed      *metrics.Int64Metric
+	mPercentUsed    *metrics.Float64Metric
 
 	config *ssmtypes.DiskStatsConfig
 
@@ -150,6 +151,17 @@ func NewDiskCollectorOrDie(diskConfig *ssmtypes.DiskStatsConfig) *diskCollector
 		klog.Fatalf("Error initializing metric for %q: %v", metrics.DiskBytesUsedID, err)
 	}
 
+	dc.mPercentUsed, err = metrics.NewFloat64Metric(
+		metrics.DiskPercentUsedID,
+		diskConfig.MetricsConfigs[string(metrics.DiskPercentUsedID)].DisplayName,
+		"Disk usage in percentage of total space",
+		"%",
+		metrics.LastValue,
+		[]string{deviceNameLabel})
+	if err != nil {
+		klog.Fatalf("Error initializing metric for %q: %v", metrics.DiskPercentUsedID, err)
+	}
+
 	dc.lastIOTime = make(map[string]uint64)
 	dc.lastWeightedIO = make(map[string]uint64)
 	dc.lastReadCount = make(map[string]uint64)
@@ -291,6 +303,9 @@ func (dc *diskCollector) collect() {
 		opttypes := strings.Join(partition.Opts, ",")
 		dc.mBytesUsed.Record(map[string]string{deviceNameLabel: deviceName, fsTypeLabel: fstype, mountOptionLabel: opttypes, stateLabel: "free"}, int64(usageStat.Free))
 		dc.mBytesUsed.Record(map[string]string{deviceNameLabel: deviceName, fsTypeLabel: fstype, mountOptionLabel: opttypes, stateLabel: "used"}, int64(usageStat.Used))
+		if dc.mPercentUsed != nil {
+			dc.mPercentUsed.Record(map[string]string{deviceNameLabel: deviceName, fsTypeLabel: fstype, mountOptionLabel: opttypes, stateLabel: "used"}, float64(usageStat.UsedPercent))
+		}
 	}
 
 }

diff --git a/pkg/systemstatsmonitor/memory_collector.go b/pkg/systemstatsmonitor/memory_collector.go
@@ -25,6 +25,7 @@ import (
 
 type memoryCollector struct {
 	mBytesUsed       *metrics.Int64Metric
+	mPercentUsed     *metrics.Float64Metric
 	mAnonymousUsed   *metrics.Int64Metric
 	mPageCacheUsed   *metrics.Int64Metric
 	mUnevictableUsed *metrics.Int64Metric
@@ -49,6 +50,17 @@ func NewMemoryCollectorOrDie(memoryConfig *ssmtypes.MemoryStatsConfig) *memoryCo
 		klog.Fatalf("Error initializing metric for %q: %v", metrics.MemoryBytesUsedID, err)
 	}
 
+	mc.mPercentUsed, err = metrics.NewFloat64Metric(
+		metrics.MemoryPercentUsedID,
+		memoryConfig.MetricsConfigs[string(metrics.MemoryPercentUsedID)].DisplayName,
+		"Memory usage in percentage of total memory.",
+		"%",
+		metrics.LastValue,
+		[]string{stateLabel})
+	if err != nil {
+		klog.Fatalf("Error initializing metric for %q: %v", metrics.MemoryPercentUsedID, err)
+	}
+
 	mc.mAnonymousUsed, err = metrics.NewInt64Metric(
 		metrics.MemoryAnonymousUsedID,
 		memoryConfig.MetricsConfigs[string(metrics.MemoryAnonymousUsedID)].DisplayName,

diff --git a/pkg/systemstatsmonitor/memory_collector_unix.go b/pkg/systemstatsmonitor/memory_collector_unix.go
@@ -58,6 +58,12 @@ func (mc *memoryCollector) collect() {
 		}
 	}
 
+	if mc.mPercentUsed != nil && meminfo.MemTotal != nil && *meminfo.MemTotal > 0 &&
+		meminfo.MemFree != nil && meminfo.Buffers != nil && meminfo.Cached != nil && meminfo.Slab != nil {
+		ratio := float64(*meminfo.MemTotal - *meminfo.MemFree - *meminfo.Buffers - *meminfo.Cached - *meminfo.Slab) / float64(*meminfo.MemTotal)
+		mc.mPercentUsed.Record(map[string]string{stateLabel: "used"}, float64(ratio*100.0))
+	}
+
 	if mc.mDirtyUsed != nil {
 		if meminfo.Dirty != nil {
 			mc.mDirtyUsed.Record(map[string]string{stateLabel: "dirty"}, int64(*meminfo.Dirty)*1024)

diff --git a/pkg/systemstatsmonitor/memory_collector_windows.go b/pkg/systemstatsmonitor/memory_collector_windows.go
@@ -37,4 +37,8 @@ func (mc *memoryCollector) collect() {
 		mc.mBytesUsed.Record(map[string]string{stateLabel: "free"}, int64(meminfo.Available)*1024)
 		mc.mBytesUsed.Record(map[string]string{stateLabel: "used"}, int64(meminfo.Used)*1024)
 	}
+
+	if mc.mPercentUsed != nil {
+		mc.mPercentUsed.Record(map[string]string{stateLabel: "used"}, float64(meminfo.UsedPercent))
+	}
 }
diff --git a/pkg/util/metrics/metric.go b/pkg/util/metrics/metric.go
@@ -35,12 +35,14 @@ const (
 	DiskOpsBytesID          MetricID = "disk/operation_bytes_count"
 	DiskOpsTimeID           MetricID = "disk/operation_time"
 	DiskBytesUsedID         MetricID = "disk/bytes_used"
+	DiskPercentUsedID       MetricID = "disk/percent_used"
 	HostUptimeID            MetricID = "host/uptime"
 	MemoryBytesUsedID       MetricID = "memory/bytes_used"
 	MemoryAnonymousUsedID   MetricID = "memory/anonymous_used"
 	MemoryPageCacheUsedID   MetricID = "memory/page_cache_used"
 	MemoryUnevictableUsedID MetricID = "memory/unevictable_used"
 	MemoryDirtyUsedID       MetricID = "memory/dirty_used"
+	MemoryPercentUsedID     MetricID = "memory/percent_used"
 	OSFeatureID             MetricID = "system/os_feature"
 	SystemProcessesTotal    MetricID = "system/processes_total"
 	SystemProcsRunning      MetricID = "system/procs_running"

diff --git a/test/e2e/metriconly/metrics_test.go b/test/e2e/metriconly/metrics_test.go
@@ -85,13 +85,15 @@ var _ = ginkgo.Describe("NPD should export Prometheus metrics.", func() {
 			assertMetricExist(gotMetrics, "disk_operation_bytes_count", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "disk_operation_time", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "disk_bytes_used", map[string]string{}, false)
+			assertMetricExist(gotMetrics, "disk_percent_used", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "disk_io_time", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "disk_weighted_io", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "memory_bytes_used", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "memory_anonymous_used", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "memory_page_cache_used", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "memory_unevictable_used", map[string]string{}, true)
 			assertMetricExist(gotMetrics, "memory_dirty_used", map[string]string{}, false)
+			assertMetricExist(gotMetrics, "memory_percent_used", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "host_uptime", map[string]string{}, false)
 			assertMetricExist(gotMetrics, "system_os_feature", map[string]string{}, false)
 		})