d.Graph(
title="coredns: # running",
targets=[
d.Target(
expr=
'count(container_memory_usage_bytes{namespace="kube-system", container="coredns"}) by (container, namespace)'
)
],
nullPointMode="null",
),
d.Graph(
title="coredns: memory usage",
targets=d.min_max_avg(
base=
'process_resident_memory_bytes{namespace="kube-system", job="kube-dns"}',
by=["job", "namespace"],
legend="{{job}}",
),
nullPointMode="null",
yAxes=g.single_y_axis(format=g.BYTES_FORMAT),
),
]
dashboard = d.Dashboard(
title="DNS",
rows=[
d.Row(title="In-cluster DNS latency", panels=DNS_LATENCY_PANEL),
d.Row(title="CoreDNS", panels=COREDNS_PANELS),
],
).auto_panel_ids()
],
yAxes=g.single_y_axis(format=g.SHORT_FORMAT, logBase=10),
),
]
# The final dashboard must be named 'dashboard' so that grafanalib will find it.
dashboard = d.Dashboard(
title="Master dashboard",
refresh="",
rows=[
d.Row(title="Clusterloader", panels=CLUSTERLOADER_PANELS),
d.Row(title="Overall cluster health",
panels=HEALTH_PANELS,
collapse=True),
d.Row(title="etcd", panels=ETCD_PANELS, collapse=True),
d.Row(title="kube-apiserver", panels=APISERVER_PANELS, collapse=True),
d.Row(
title="kube-controller-manager",
panels=[
d.simple_graph(
"Workqueue depths",
'workqueue_depth{endpoint="kube-controller-manager"}',
legend="{{name}}",
)
],
collapse=True,
),
d.Row(title="Master VM", panels=VM_PANELS, collapse=True),
],
).auto_panel_ids()
d.Graph(
title="Service: # running",
targets=[
d.Target(
expr=
'count(process_resident_memory_bytes{namespace="kube-system", job="kube-dns"}) by (job, namespace)'
)
],
nullPointMode="null",
),
d.Graph(
title="Service: memory usage",
targets=d.min_max_avg(
base=
'process_resident_memory_bytes{namespace="kube-system", job="kube-dns"}',
by=["job", "namespace"],
legend="{{job}}",
),
nullPointMode="null",
yAxes=g.single_y_axis(format=g.BYTES_FORMAT),
),
]
dashboard = d.Dashboard(
title="DNS",
rows=[
d.Row(title="In-cluster DNS prober", panels=PROBER_PANEL),
d.Row(title="In-cluster DNS service", panels=SERVICE_PANELS),
],
).auto_panel_ids()
dashboard = d.Dashboard(
title="Master dashboard",
rows=[
d.Row(title="Clusterloader", panels=CLUSTERLOADER_PANELS),
d.Row(title="Overall cluster health", panels=HEALTH_PANELS),
d.Row(title="etcd", panels=ETCD_PANELS),
d.Row(title="kube-apiserver", panels=APISERVER_PANELS),
d.Row(
title="kube-controller-manager",
panels=[
d.simple_graph(
"Workqueue depths",
'workqueue_depth{endpoint="kube-controller-manager"}',
legend="{{name}}",
)
],
),
d.Row(title="Master VM", panels=VM_PANELS),
d.Row(
title="Addons",
panels=[
d.Graph(
title="Coredns memory",
dataSource="$source",
targets=[
g.Target(
expr=
'quantile(1, sum(process_resident_memory_bytes{job="kube-dns"}) by (pod))',
legendFormat="coredns-mem-100pctl",
),
g.Target(
expr=
'quantile(0.99, sum(process_resident_memory_bytes{job="kube-dns"}) by (pod))',
legendFormat="coredns-mem-99ctl",
),
g.Target(
expr=
'quantile(0.90, sum(process_resident_memory_bytes{job="kube-dns"}) by (pod))',
legendFormat="coredns-mem-90ctl",
),
g.Target(
expr=
'quantile(0.50, sum(process_resident_memory_bytes{job="kube-dns"}) by (pod))',
legendFormat="coredns-mem-50ctl",
),
],
yAxes=g.single_y_axis(format=g.BYTES_FORMAT),
)
],
),
],
).auto_panel_ids()
dashboard = d.Dashboard(
title="Comparison Master dashboard",
refresh="",
rows=[
d.Row(title="API call latency", panels=extended_copy(API_CALL_LATENCY_PANELS)),
d.Row(title="API call latency aggregated with quantile", panels=extended_copy(QUANTILE_API_CALL_LATENCY_PANELS), collapse=True),
d.Row(title="P&F metrics", panels=extended_copy(PAF_PANELS), collapse=True),
d.Row(title="Overall cluster health", panels=extended_copy(HEALTH_PANELS), collapse=True),
d.Row(title="etcd", panels=extended_copy(ETCD_PANELS), collapse=True),
d.Row(title="kube-apiserver", panels=extended_copy(APISERVER_PANELS), collapse=True),
d.Row(title="kube-controller-manager", panels=extended_copy(CONTROLLER_MANAGER_PANELS), collapse=True),
d.Row(title="Master VM", panels=extended_copy(VM_PANELS), collapse=True),
],
templating=g.Templating(
list=[
d.SOURCE_TEMPLATE,
g.Template(
name="secondary_source",
type="datasource",
query="prometheus",
),
g.Template(
name="timeshift",
type="interval",
query="",
),
g.Template(
name="etcd_type",
type="query",
dataSource="$source",
regex=r"\*\[+\]+(.*)",
query="label_values(etcd_request_duration_seconds_count, type)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="etcd_operation",
type="query",
dataSource="$source",
query="label_values(etcd_request_duration_seconds_count, operation)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="verb",
type="query",
dataSource="$source",
query="label_values(apiserver_request_duration_seconds_count, verb)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="resource",
type="query",
dataSource="$source",
regex="(.*)s",
query="label_values(apiserver_request_duration_seconds_count, resource)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
]
),
).auto_panel_ids()
dashboard = d.Dashboard(
title="Master dashboard",
refresh="",
rows=[
d.Row(title="API call latency", panels=API_CALL_LATENCY_PANELS),
d.Row(title="API call latency aggregated with quantile",
panels=QUANTILE_API_CALL_LATENCY_PANELS,
collapse=True),
d.Row(title="Overall cluster health",
panels=HEALTH_PANELS,
collapse=True),
d.Row(title="etcd", panels=ETCD_PANELS, collapse=True),
d.Row(title="kube-apiserver", panels=APISERVER_PANELS, collapse=True),
d.Row(
title="kube-controller-manager",
panels=[
d.simple_graph(
"Workqueue depths",
'workqueue_depth{endpoint="kube-controller-manager"}',
legend="{{name}}",
)
],
collapse=True,
),
d.Row(title="Master VM", panels=VM_PANELS, collapse=True),
],
templating=g.Templating(list=[
d.SOURCE_TEMPLATE,
g.Template(
name="etcd_type",
type="query",
dataSource="$source",
regex=r"\*\[+\]+(.*)",
query="label_values(etcd_request_duration_seconds_count, type)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="etcd_operation",
type="query",
dataSource="$source",
query=
"label_values(etcd_request_duration_seconds_count, operation)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="verb",
type="query",
dataSource="$source",
query=
"label_values(apiserver_request_duration_seconds_count, verb)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
g.Template(
name="resource",
type="query",
dataSource="$source",
regex="(.*)s",
query=
"label_values(apiserver_request_duration_seconds_count, resource)",
multi=True,
includeAll=True,
refresh=g.REFRESH_ON_TIME_RANGE_CHANGE,
),
]),
).auto_panel_ids()
api_call_latency(
title="Read-only API call latency (scope=cluster, threshold=30s)",
metric=metric,
verb="LIST",
scope="cluster",
threshold=30,
),
api_call_latency(
title="Mutating API call latency (threshold=1s)",
metric=metric,
verb=d.any_of("CREATE", "DELETE", "PATCH", "POST", "PUT"),
scope=d.any_of("namespace", "cluster"),
threshold=1,
),
]
# The final dashboard must be named 'dashboard' so that grafanalib will find it.
dashboard = d.Dashboard(
title="SLO",
rows=[
d.Row(title="SLO", panels=create_slo_panel()),
d.Row(
title="Experimental: SLO (window 1m)",
panels=create_slo_panel(
metric="apiserver:apiserver_request_latency_1m:histogram_quantile"
),
),
],
).auto_panel_ids()
nullPointMode="null",
),
d.Graph(
title="probes: memory usage",
targets=[
d.Target(
expr='min(container_memory_usage_bytes{namespace="probes", container=~"ping-client|ping-server"}) by (container)',
legendFormat="min {{container}}",
),
d.Target(
expr='avg(container_memory_usage_bytes{namespace="probes", container=~"ping-client|ping-server"}) by (container)',
legendFormat="avg {{container}}",
),
d.Target(
expr='max(container_memory_usage_bytes{namespace="probes", container=~"ping-client|ping-server"}) by (container)',
legendFormat="max {{container}}",
),
],
nullPointMode="null",
),
]
dashboard = d.Dashboard(
title="Network",
rows=[
d.Row(title="Network progamming latency", panels=NETWORK_PROGRAMMING_PANEL),
d.Row(title="In-cluster network latency", panels=NETWORK_LATENCY_PANEL),
],
).auto_panel_ids()
