Collect Kubernetes metrics and logs with Splunk App for Infrastructure

Collect metrics and log data from a Kubernetes cluster with the easy install script in the Splunk App for Infrastructure (SAI). When you run the script, you start ingesting metrics and log data for pods and nodes in the cluster. Nodes and pods in the cluster you monitor are entities in SAI. You can search other metrics you specify to collect data for in the Search app.

SAI deploys Splunk Connect for Kubernetes (SCK) with Helm to collect metrics and log data from Kubernetes clusters. This version of SAI deploys SCK version 1.3.0. For more information about SCK, see the Splunk Connect for Kubernetes 1.3.0 release documentation in the Github repository.

View detailed information about the status of pods you monitor from the Entity Overview. For information about pod statuses, see Pod phase on the Kubernetes website. The status for Kubernetes nodes is set to disabled when the status of then node enters an unknown state. From the Investigate tab, the status of entities does not contain detailed pod status information, and is either Active or Inactive.

Go to the Investigate page in SAI to monitor your entities in the Tile or List view. You can group your entities to monitor them more easily, and further analyze your infrastructure by drilling down to the Overview Dashboard for entities or Analysis Workspace for entities and groups.

For information about stopping or removing the data collection agents, see Stop data collection on Splunk App for Infrastructure.

Prerequisites

Meet the following requirements to configure data collection:

Item	Requires
Data collection script dependencies	See Kubernetes data collection requirements in the Install and Upgrade Splunk App for Infrastructure guide.
Helm	You must have permission to execute helm commands.
HEC token	To configure an HEC token for SAI, see Configure the HTTP Event Collector to receive metrics data for SAI.

Steps

Follow these steps to configure and run the data collection script to start forwarding data from a Kubernetes cluster to SAI.

1. Set up Helm

Install and initialize Helm on each Kubernetes cluster you want to monitor in SAI. For information about setting up Helm, see the Quickstart Guide on the Helm website.

You must run the easy install script on the system that runs Helm.

2. Specify configuration options

Specify the data collection options for collecting metrics and logs from the cluster. If you're running SAI on Splunk Cloud, you must enter specific settings for the Monitoring machine, HEC port, and Receiver port. For more information, see Install and configure the data collection agents on each applicable system in the Install and Upgrade Splunk App for Infrastructure guide.

1. In the SAI user interface, click the Add Data tab and select Kubernetes.
2. For Data to be collected, click Customize Objects to define which objects to track:

   Object	Sourcetype	Description
   pods	kube:objects:pods	Enabled by default, cannot be disabled. Collects metadata, spec, and status data for pods in the cluster.
   nodes	kube:objects:nodes	Enabled by default, and cannot be disabled. Collects metadata, spec, and status data for nodes in the cluster.

   You can enable advanced object collection for these objects:

   Object	Sourcetype	Description
   component_statuses	kube:objects:component_statuses	Collects conditions and metadata data for the status of resources in the cluster.
   config_maps	kube:objects:config_maps	Collects data and metadata data for ConfigMaps in the cluster.
   daemon_sets	kube:objects:daemon_sets	Collects metadata, spec, and status data for daemonsets in the cluster.
   deployments	kube:objects:deployments	Collects metadata, spec, and status data for deployments in the cluster.
   namespaces	kube:objects:namespaces	Collects metadata, spec, and status data for namespaces in the cluster.
   persistent_volumes	kube:objects:persistent_volumes	Collects metadata, spec, and status data for persistent volumes in the cluster.
   persistent_volume_claims	kube:objects:persistent_volume_claims	Collects metadata, spec, and status data for persistent volume claims in the cluster.
   replica_sets	kube:objects:replica_sets	Collects metadata, spec, and status data for replica sets in the cluster.
   resource_quotas	kube:objects:resource_quotas	Collects metadata, spec, and status data for resource quotas in the cluster.
   services	kube:objects:services	Collects metadata, spec, and status data for services in the cluster.
   service_accounts	kube:objects:service_accounts	Collects metadata and secrets data for service accounts in the cluster.
   stateful_sets	kube:objects:stateful_sets	Collects metadata, spec, and status data for stateful sets in the cluster.
   events	kube:objects:events	Collects object and type data for events in the cluster.

   Advanced object collection options do not have visualizations in SAI. Track these objects in the Search & Reporting app. By default, object data is stored in the em_meta index.

3. For Monitoring machine, enter the FQDN or IP address of the system you are sending data to. This is the system running SAI.
4. Enter the HEC token of the system you want to send data to.
5. Enter the HEC port of the system you want to send metrics data to. Use port 8088 if it is available.
6. Enter a unique Kubernetes namespace to specify a namespace in the Kubernetes cluster for the SCK components.
7. Enter a unique Cluster name to specify the name of the Kubernetes cluster you're running the script in. If you do not enter anything, the script specifies a name for you.
8. Enter a unique Release name for the SCK release when you install it in your cluster. The release tracks the installation of SCK.

3. Run the script

Execute the script on the system that runs Helm.

1. Open a command line window on the system that runs Helm.
2. Switch to the cluster you want to monitor in SAI:

   kubectl config use-context <context_name>
   

   where <context_name> is the context that corresponds to the cluster you're monitoring.
3. Paste the script you configured in the Add Data tab in SAI.
4. To verify you successfully deployed the SCK, check the status of the release in the console using $ helm status <release name> from the command line.

Upgrade Splunk Connect for Kubernetes in SAI

You can deploy a more recent version of SCK to a Kubernetes cluster you're already monitoring. If you started collecting Kubernetes data with an earlier version of SAI, you may be running an earlier version of SCK. To upgrade SCK, delete SCK from the cluster and then run the data collection script from a more recent version of SAI that deploys a more recent version of SCK. When you upgrade SCK, SAI discovers resources in the cluster as new entities.

When you delete SCK from the cluster during the upgrade process, you can manually delete entities associated with the cluster or wait for your Kubernetes entity retirement policy to automatically remove the entities. If you don't have a Kubernetes entity retirement policy and don't manually delete the entities for a cluster after you upgrade SCK in the cluster, the old entities that the earlier version of SCK discovered just become inactive.

This version of SAI deploys SCK version 1.3.0 when you run the data collection script. For more information about SCK, see the Splunk Connect for Kubernetes 1.3.0 release documentation in the Github repository.

Follow these steps to upgrade SCK:

1. On the system that runs SCK, delete the Helm release name for the current SCK deployment:

   $ helm delete <helm_release_name> --purge

2. Delete entities that the version of SCK you're replacing discovered. Alternatively, you can wait for an active Kubernetes entity retirement policy to automatically remove the entities when they become inactive.
3. Configure the data collection script in a version of SAI that deploys a more recent version of SCK.