File tree Expand file tree Collapse file tree 3 files changed +94
-0
lines changed Expand file tree Collapse file tree 3 files changed +94
-0
lines changed Original file line number Diff line number Diff line change
1
+ {{- if .Values.devicePlugin.serviceMonitor.enabled }}
2
+ apiVersion : monitoring.coreos.com/v1
3
+ kind : ServiceMonitor
4
+ metadata :
5
+ name : {{ include "hami-vgpu.device-plugin" . }}
6
+ namespace : {{ include "hami-vgpu.namespace" . }}
7
+ labels :
8
+ release : prometheus
9
+ spec :
10
+ endpoints :
11
+ - bearerTokenSecret :
12
+ key : " "
13
+ name : " "
14
+ interval : 15s
15
+ metricRelabelings :
16
+ - action : replace
17
+ sourceLabels :
18
+ - podname
19
+ targetLabel : exported_pod
20
+ - action : replace
21
+ sourceLabels :
22
+ - podnamespace
23
+ targetLabel : exported_namespace
24
+ - targetLabel : cluster
25
+ replacement : local-cluster
26
+ - sourceLabels : [podname]
27
+ regex : ^(.+)-[a-z0-9]{9,}-[a-z0-9]+$
28
+ targetLabel : deployment_name
29
+ replacement : $1
30
+ action : replace
31
+ - sourceLabels : [podnamespace, deployment_name]
32
+ separator : /
33
+ targetLabel : deployment
34
+ action : replace
35
+ path : /metrics
36
+ port : monitorport
37
+ relabelings :
38
+ - action : replace
39
+ sourceLabels :
40
+ - __meta_kubernetes_endpoint_node_name
41
+ targetLabel : node
42
+ jobLabel : app
43
+ namespaceSelector :
44
+ matchNames :
45
+ - {{ include "hami-vgpu.namespace" . }}
46
+ selector :
47
+ matchLabels :
48
+ app.kubernetes.io/component : hami-device-plugin
49
+ {{- end }}
Original file line number Diff line number Diff line change
1
+ {{- if .Values.scheduler.serviceMonitor.enabled }}
2
+ apiVersion : monitoring.coreos.com/v1
3
+ kind : ServiceMonitor
4
+ metadata :
5
+ name : {{ include "hami-vgpu.scheduler" . }}
6
+ namespace : {{ include "hami-vgpu.namespace" . }}
7
+ labels :
8
+ release : prometheus
9
+ spec :
10
+ endpoints :
11
+ - bearerTokenSecret :
12
+ key : " "
13
+ name : " "
14
+ interval : 15s
15
+ metricRelabelings :
16
+ - action : replace
17
+ sourceLabels :
18
+ - podname
19
+ targetLabel : exported_pod
20
+ - action : replace
21
+ sourceLabels :
22
+ - podnamespace
23
+ targetLabel : exported_namespace
24
+ - action : replace
25
+ sourceLabels :
26
+ - nodeid
27
+ targetLabel : node
28
+ - targetLabel : cluster
29
+ replacement : local-cluster
30
+ path : /metrics
31
+ port : monitor
32
+ jobLabel : app
33
+ namespaceSelector :
34
+ matchNames :
35
+ - {{ include "hami-vgpu.namespace" . }}
36
+ selector :
37
+ matchLabels :
38
+ app.kubernetes.io/component : hami-scheduler
39
+ {{- end }}
Original file line number Diff line number Diff line change @@ -232,6 +232,9 @@ scheduler:
232
232
monitorTargetPort : 9395
233
233
labels : {}
234
234
annotations : {}
235
+ # enable serviceMonitor resource for hami scheduler
236
+ serviceMonitor :
237
+ enabled : false
235
238
236
239
devicePlugin :
237
240
# # @param image.registry devicePlugin image registry
@@ -296,6 +299,9 @@ devicePlugin:
296
299
httpPort : 31992
297
300
labels : {}
298
301
annotations : {}
302
+ # enable serviceMonitor resource for device plugin
303
+ serviceMonitor :
304
+ enabled : false
299
305
300
306
pluginPath : /var/lib/kubelet/device-plugins
301
307
libPath : /usr/local/vgpu
You can’t perform that action at this time.
0 commit comments