9cc4786a9 · cryptotaxi247/netdata-1

@@ -1,137 +1,42 @@

1 -<!--

2 -title: "Kubernetes visualizations"

3 -description: "Netdata Cloud features rich, zero-configuration Kubernetes monitoring for the resource utilization and application metrics of Kubernetes (k8s) clusters."

4 -custom_edit_url: "https://github.com/netdata/netdata/blob/master/docs/cloud/visualize/kubernetes.md"

5 -sidebar_label: "Kubernetes visualizations"

6 -learn_status: "Published"

7 -learn_topic_type: "Concepts"

8 -learn_rel_path: "Operations/Visualizations"

9 --->

1 +# Kubernetes tab

2

11 -# Kubernetes visualizations

3 +The Netdata dashboards feature enhanced visualizations for the resource utilization of Kubernetes (k8s) clusters, embedded in the default [Metrics tab](https://github.com/netdata/netdata/blob/master/docs/dashboard/metrics-tab-and-single-node-tabs.md) dashboard.

4

13 -Netdata Cloud features enhanced visualizations for the resource utilization of Kubernetes (k8s) clusters, embedded in the default [Metrics tab](https://github.com/netdata/netdata/blob/master/docs/dashboard/metrics-tab-and-single-node-tabs.md) dashboard.

5 +These visualizations include a health map for viewing the status of k8s pods/containers, in addition to [Netdata charts](https://github.com/netdata/netdata/blob/master/docs/cloud/visualize/interact-new-charts.md) for viewing per-second CPU, memory, disk, and networking metrics from k8s nodes.

6

15 -These visualizations include a health map for viewing the status of k8s pods/containers, in addition to composite charts

16 -for viewing per-second CPU, memory, disk, and networking metrics from k8s nodes.

17 -

18 -See our [Kubernetes deployment instructions](https://github.com/netdata/netdata/blob/master/packaging/installer/methods/kubernetes.md) for details on

19 -installation and connecting to Netdata Cloud.

7 +See our [Kubernetes deployment instructions](https://github.com/netdata/netdata/blob/master/packaging/installer/methods/kubernetes.md) for details on deploying Netdata on your Kubernetes cluster.

8

9 ## Available Kubernetes metrics

10

11 Netdata Cloud organizes and visualizes the following metrics from your Kubernetes cluster from every container:

12

25 -- `cpu_limit`: CPU utilization as a percentage of the limit defined by the [pod specification

26 - `spec.containers[].resources.limits.cpu`](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#resource-requests-and-limits-of-pod-and-container)

27 - or a [`LimitRange`

28 - object](https://kubernetes.io/docs/tasks/administer-cluster/manage-resources/cpu-default-namespace/#create-a-limitrange-and-a-pod).

29 -- `cpu`: CPU utilization of the pod/container. 100% usage equals 1 fully-utilized core, 200% equals 2 fully-utilized

30 - cores, and so on.

31 -- `cpu_per_core`: CPU utilization averaged across available cores.

32 -- `mem_usage_limit`: Memory utilization, without cache, as a percentage of the limit defined by the [pod specification

33 - `spec.containers[].resources.limits.memory`](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#resource-requests-and-limits-of-pod-and-container)

34 - or a [`LimitRange`

35 - object](https://kubernetes.io/docs/tasks/administer-cluster/manage-resources/cpu-default-namespace/#create-a-limitrange-and-a-pod).

36 -- `mem_usage`: Used memory, without cache.

37 -- `mem`: The sum of `cache` and `rss` (resident set size) memory usage.

38 -- `writeback`: The size of `dirty` and `writeback` cache.

39 -- `mem_activity`: Sum of `in` and `out` bandwidth.

40 -- `pgfaults`: Sum of page fault bandwidth, which are raised when the Kubernetes cluster tries accessing a memory page

41 - that is mapped into the virtual address space, but not actually loaded into main memory.

42 -- `throttle_io`: Sum of `read` and `write` per second across all PVs/PVCs attached to the container.

43 -- `throttle_serviced_ops`: Sum of the `read` and `write` operations per second across all PVs/PVCs attached to the

44 - container.

45 -- `net.net`: Sum of `received` and `sent` bandwidth per second.

46 -- `net.packets`: Sum of `multicast`, `received`, and `sent` packets.

47 -

48 -When viewing the [health map](#health-map), Netdata Cloud shows the above metrics per container, or aggregated based on

13 +| Metric | Description |

14 +|------------------------------------|-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------|

15 +| `k8s.cgroup.cpu_limit` | CPU utilization as a percentage of the limit defined by the [pod specification `spec.containers[].resources.limits.cpu`](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#resource-requests-and-limits-of-pod-and-container) or a [`LimitRange` object](https://kubernetes.io/docs/tasks/administer-cluster/manage-resources/cpu-default-namespace/#create-a-limitrange-and-a-pod). |

16 +| `k8s.cgroup.cpu` | CPU utilization of the pod/container. 100% usage equals 1 fully-utilized core, 200% equals 2 fully-utilized cores, and so on. |

17 +| `k8s.cgroup.throttled` | The percentage of runnable periods when tasks in a cgroup have been throttled. |

18 +| `k8s.cgroup.throttled_duration` | The total time duration for which tasks in a cgroup have been throttled. |

19 +| `k8s.cgroup.mem_utilization` | Memory utilization within the configured or system-wide (if not set) limits. |

20 +| `k8s.cgroup.mem_usage_limit` | Memory utilization, without cache, as a percentage of the limit defined by the [pod specification `spec.containers[].resources.limits.memory`](https://kubernetes.io/docs/concepts/configuration/manage-resources-containers/#resource-requests-and-limits-of-pod-and-container) or a [`LimitRange` object](https://kubernetes.io/docs/tasks/administer-cluster/manage-resources/cpu-default-namespace/#create-a-limitrange-and-a-pod). |

21 +| `k8s.cgroup.mem_usage` | Used memory, without cache. |

22 +| `k8s.cgroup.mem` | The sum of `cache` and `rss` (resident set size) memory usage. |

23 +| `k8s.cgroup.writeback` | The size of `dirty` and `writeback` cache. |

24 +| `k8s.cgroup.pgfaults` | Sum of page fault bandwidth, which are raised when the Kubernetes cluster tries accessing a memory page that is mapped into the virtual address space, but not actually loaded into main memory. |

25 +| `k8s.cgroup.throttle_io` | Sum of `read` and `write` per second across all PVs/PVCs attached to the container. |

26 +| `k8s.cgroup.throttle_serviced_ops` | Sum of the `read` and `write` operations per second across all PVs/PVCs attached to the container. |

27 +| `k8s.cgroup.net_net` | Sum of `received` and `sent` bandwidth per second. |

28 +| `k8s.cgroup.net_packets` | Sum of `multicast`, `received`, and `sent` packets. |

29 +

30 +

31 +When viewing the [overview of this dashboard](#kubernetes-containers-overview), Netdata presents the above metrics per container, or aggregated based on

32 their associated pods.

33

51 -When viewing the [composite charts](#composite-charts), Netdata Cloud aggregates metrics from multiple nodes, pods, or

52 -containers, depending on the grouping chosen. For example, if you group the `cpu_limit` composite chart by

53 -`k8s_namespace`, the metrics shown will be the average of `cpu_limit` metrics from all nodes/pods/containers that are

54 -part of that namespace.

55 -

56 -## Health map

57 -

58 -The health map places each container or pod as a single box, then varies the intensity of its color to visualize the

59 -resource utilization of specific k8s pods/containers.

60 -

61 -![The Kubernetes health map in Netdata

62 -Cloud](https://user-images.githubusercontent.com/1153921/106964367-39f54100-66ff-11eb-888c-5a04f8abb3d0.png)

63 -

64 -Change the health map's coloring, grouping, and displayed nodes to customize your experience and learn more about the

65 -status of your k8s cluster.

34 +## Kubernetes Containers overview

35

67 -### Color by

36 +At the top of the Kubernetes containers section there is a map, that with a given context colorizes the containers in terms of their utilization.

37

69 -Color the health map by choosing an aggregate function to apply to an [available Kubernetes

70 -metric](#available-kubernetes-metrics), then whether you to display boxes for individual pods or containers.

71 -

72 -The default is the _average, of CPU within the configured limit, organized by container_.

73 -

74 -### Group by

75 -

76 -Group the health map by the `k8s_cluster_id`, `k8s_controller_kind`, `k8s_controller_name`, `k8s_kind`, `k8s_namespace`,

77 -and `k8s_node_name`. The default is `k8s_controller_name`.

78 -

79 -### Filtering

80 -

81 -Filtering behaves identically to the [node filter in War Rooms](https://github.com/netdata/netdata/blob/master/docs/cloud/visualize/node-filter.md), with the ability to

82 -filter pods/containers by `container_id` and `namespace`.

38 +The filtering of this map is controlled by using the [NIDL framework](https://github.com/netdata/netdata/blob/master/docs/cloud/visualize/interact-new-charts.md#nidl-framework) from the definition bar of the chart.

39

40 ### Detailed information

41

86 -Hover over any of the pods/containers in the map to display a modal window, which contains contextual information

87 -and real-time metrics from that resource.

88 -

89 -![The modal containing additional information about a k8s

90 -resource](https://user-images.githubusercontent.com/1153921/106964369-3a8dd780-66ff-11eb-8a8a-a5c8f0d5711f.png)

91 -

92 -The **context** tab provides the following details about a container or pod:

93 -

94 -- Cluster ID

95 -- Node

96 -- Controller Kind

97 -- Controller Name

98 -- Pod Name

99 -- Container

100 -- Kind

101 -- Pod UID

102 -

103 -This information helps orient you as to where the container/pod operates inside your cluster.

104 -

105 -The **Metrics** tab contains charts visualizing the last 15 minutes of the same metrics available in the [color by

106 -option](#color-by). Use these metrics along with the context, to identify which containers or pods are experiencing

107 -problematic behavior to investigate further, troubleshoot, and remediate with `kubectl` or another tool.

108 -

109 -## Composite charts

110 -

111 -The Kubernetes composite charts show real-time and historical resource utilization metrics from nodes, pods, or

112 -containers within your Kubernetes deployment.

113 -

114 -Kubernetes composite charts can also be

115 -grouped by the following labels:

116 -

117 -- `k8s_cluster_id`

118 -- `k8s_container_id`

119 -- `k8s_container_name`

120 -- `k8s_controller_kind`

121 -- `k8s_kind`

122 -- `k8s_namespace`

123 -- `k8s_node_name`

124 -- `k8s_pod_name`

125 -- `k8s_pod_uid`

126 -

127 -In addition, when you hover over a composite chart, the colors in the heat map changes as well, so you can see how

128 -certain pod/container-level metrics change over time.

129 -

130 -## Caveats

131 -

132 -There are some caveats and known issues with Kubernetes monitoring with Netdata Cloud.

133 -

134 -- **No way to remove any nodes** you might have

135 - [drained](https://kubernetes.io/docs/tasks/administer-cluster/safely-drain-node/) from your Kubernetes cluster. These

136 - drained nodes will be marked "unreachable" and will show up in War Room management screens/dropdowns. The same applies

137 - for any ephemeral nodes created and destroyed during horizontal scaling.

42 +Hover over any of the pods/containers in the map to display a modal window, which contains contextual information and real-time metrics from that resource.

k8s doc edits (#17471)