Add Misc cgroup support to Prometheus metrics

google · Jan 8, 2024 · 1c832cf · 1c832cf
1 parent b0c1fb5
commit 1c832cf
Show file tree

Hide file tree

Showing 5 changed files with 65 additions and 0 deletions.
diff --git a/docs/storage/prometheus.md b/docs/storage/prometheus.md
@@ -50,6 +50,8 @@ Metric name | Type | Description | Unit (where applicable) | option parameter |
 `container_hugetlb_failcnt` | Counter | Number of hugepage usage hits limits | | hugetlb |
 `container_hugetlb_max_usage_bytes` | Gauge | Maximum hugepage usages recorded | bytes | hugetlb |
 `container_hugetlb_usage_bytes` | Gauge | Current hugepage usage | bytes | hugetlb |
+`container_misc_usage` | Gauge | Current usage of the misc scalar resource specified by the label | | misc |
+`container_misc_events` | Counter | Number of times the usage for the misc scalar resource specified by the label was about to go over the max boundary | | misc |
 `container_last_seen` | Gauge | Last time a container was seen by the exporter | timestamp | - |
 `container_llc_occupancy_bytes` | Gauge | Last level cache usage statistics for container counted with RDT Memory Bandwidth Monitoring (MBM). | bytes | resctrl |
 `container_memory_bandwidth_bytes` | Gauge | Total memory bandwidth usage statistics for container counted with RDT Memory Bandwidth Monitoring (MBM). | bytes | resctrl |

diff --git a/metrics/prometheus.go b/metrics/prometheus.go
@@ -361,6 +361,43 @@ func NewPrometheusCollector(i infoProvider, f ContainerLabelsFunc, includedMetri
 			},
 		}...)
 	}
+	if includedMetrics.Has(container.MiscMetrics) {
+		c.containerMetrics = append(c.containerMetrics, []containerMetric{
+			{
+				name:        "container_misc_usage",
+				help:        "Current usage of the misc scalar resource specified by the label",
+				valueType:   prometheus.GaugeValue,
+				extraLabels: []string{"resource"},
+				getValues: func(s *info.ContainerStats) metricValues {
+					values := make(metricValues, 0, len(s.Misc))
+					for k, v := range s.Misc {
+						values = append(values, metricValue{
+							value:     float64(v.Usage),
+							labels:    []string{k},
+							timestamp: s.Timestamp,
+						})
+					}
+					return values
+				},
+			}, {
+				name:        "container_misc_events",
+				help:        "Number of times the usage for the misc scalar resource specified by the label was about to go over the max boundary",
+				valueType:   prometheus.CounterValue,
+				extraLabels: []string{"resource"},
+				getValues: func(s *info.ContainerStats) metricValues {
+					values := make(metricValues, 0, len(s.Misc))
+					for k, v := range s.Misc {
+						values = append(values, metricValue{
+							value:     float64(v.Events),
+							labels:    []string{k},
+							timestamp: s.Timestamp,
+						})
+					}
+					return values
+				},
+			},
+		}...)
+	}
 	if includedMetrics.Has(container.MemoryUsageMetrics) {
 		c.containerMetrics = append(c.containerMetrics, []containerMetric{
 			{

diff --git a/metrics/prometheus_fake.go b/metrics/prometheus_fake.go
@@ -368,6 +368,16 @@ func (p testSubcontainersInfoProvider) GetRequestedContainersInfo(string, v2.Req
 							Failcnt:  0,
 						},
 					},
+					Misc: map[string]info.MiscStats{
+						"res_a": {
+							Usage:  1,
+							Events: 42,
+						},
+						"res_b": {
+							Usage:  2,
+							Events: 42,
+						},
+					},
 					Network: info.NetworkStats{
 						InterfaceStats: info.InterfaceStats{
 							Name:      "eth0",

diff --git a/metrics/testdata/prometheus_metrics b/metrics/testdata/prometheus_metrics
@@ -133,6 +133,14 @@ container_hugetlb_max_usage_bytes{container_env_foo_env="prod",container_label_f
 # TYPE container_hugetlb_usage_bytes gauge
 container_hugetlb_usage_bytes{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",pagesize="1Gi",zone_name="hello"} 0 1395066363000
 container_hugetlb_usage_bytes{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",pagesize="2Mi",zone_name="hello"} 4 1395066363000
+# HELP container_misc_usage Current usage of the resource
+# TYPE container_misc_usage gauge
+container_misc_usage{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",resource="res_a",zone_name="hello"} 1 1395066363000
+container_misc_usage{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",resource="res_b",zone_name="hello"} 2 1395066363000
+# HELP container_misc_events Number of times the usage for the resource was about to go over the max boundary
+# TYPE container_misc_events counter
+container_misc_events{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",resource="res_a",zone_name="hello"} 42 1395066363000
+container_misc_events{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",resource="res_b",zone_name="hello"} 42 1395066363000
 # HELP container_last_seen Last time a container was seen by the exporter
 # TYPE container_last_seen gauge
 container_last_seen{container_env_foo_env="prod",container_label_foo_label="bar",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 1.395066363e+09 1395066363000

diff --git a/metrics/testdata/prometheus_metrics_whitelist_filtered b/metrics/testdata/prometheus_metrics_whitelist_filtered
@@ -133,6 +133,14 @@ container_hugetlb_max_usage_bytes{container_env_foo_env="prod",id="testcontainer
 # TYPE container_hugetlb_usage_bytes gauge
 container_hugetlb_usage_bytes{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",pagesize="1Gi",zone_name="hello"} 0 1395066363000
 container_hugetlb_usage_bytes{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",pagesize="2Mi",zone_name="hello"} 4 1395066363000
+# HELP container_misc_usage Current usage of the resource
+# TYPE container_misc_usage gauge
+container_misc_usage{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",resource="res_a",zone_name="hello"} 1 1395066363000
+container_misc_usage{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",resource="res_b",zone_name="hello"} 2 1395066363000
+# HELP container_misc_events Number of times the usage for the resource was about to go over the max boundary
+# TYPE container_misc_events counter
+container_misc_events{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",resource="res_a",zone_name="hello"} 42 1395066363000
+container_misc_events{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",resource="res_b",zone_name="hello"} 42 1395066363000
 # HELP container_last_seen Last time a container was seen by the exporter
 # TYPE container_last_seen gauge
 container_last_seen{container_env_foo_env="prod",id="testcontainer",image="test",name="testcontaineralias",zone_name="hello"} 1.395066363e+09 1395066363000