Azure Kubernetes Cluster Sensor
The sensor monitors Azure Kubernetes Services resource, using Azure Monitor metrics.
Azure Kubernetes Cluster Sensor allows collecting essential metrics for the managed Cluster. With this sensor you can for example be notified when the allocable memory on the Cluster is running out.
Options
To use this sensor you will need Client ID, Tenant ID, and Application Secret.
Alerts
- Used Memory Working Set > 70%
- Used Memory Working Set > 90%
- Disk Used > 70%
- Disk Used > 90%
- RSS Memory Used > 70%
- RSS Memory Used > 90%
Performance Metrics
- Container\Used Memory Working Set Bytes - Container working set memory used in bytes
- Container\% Used Memory Working Set - Container working set memory used in percent
- Container\RSS Memory Bytes - Container RSS memory used in bytes
- Container\% RSS Memory Used - Container working set memory used in percent
- Nodes\CPU Cores Allocatable - Total number of available CPU cores in a managed cluster
- Nodes\Memory Bytes Allocatable - Total amount of available memory in a managed cluster in bytes
- Nodes\Nodes Condition - Number of nodes in a known condition
- Nodes\CPU Milicores Utilization - Aggregated measurement of CPU utilization in millicores across the cluster
- Nodes\% CPU Cluster Utilization - Aggregated average CPU utilization measured in percentage across the cluster
- Nodes\Disk Used Bytes - Disk space used in bytes by device
- Nodes\% Disk Used - Disk space used in percent by device
- Nodes\Bytes Received - Network received bytes
- Nodes\Bytes Sent - Network transmitted bytes
- Cluster Autoscaler\Cluster Health - Determines whether or not cluster autoscaler will take action on the cluster.
- Cluster Autoscaler\Nodes Cooldown - Determines if the scale down is in cooldown - No nodes will be removed during this timeframe
- Cluster Autoscaler\Nodes Unneeded - Cluster autoscaler marks those nodes as candidates for deletion and are eventually deleted
- Cluster Autoscaler\Pods Unschedulable - Number of pods that are currently unschedulable in the cluster
- Pods\Phase - Number of pods in known phase
- Pods\Ready - Number of pods in Ready state
- API Server\Avg. Requests Inflight - Maximum number of currently active inflight requests on the API Server per request kind
- Sensor\Check Time - Time of the monitoring operation
Reports (Data Collectors)
By default, the sensor adds a report containing Container\Used Memory Working Set Bytes, Container\RSS Memory Bytes, Nodes\Nodes Condition, Nodes\CPU Cores Allocatable, Nodes\Memory Bytes Allocatable, Nodes\Bytes Received, Nodes\Bytes Sent, Nodes\Disk Used Bytes, Pods\Phase, and Pods\Ready charts. These metrics will be collected and available as a report or through @trend-viewer