Collected information

In order to accurately determine the pricing and utilization for each node, pod, and volume, the agent (the term "agent" here refers to the entire collection of services mentioned above) watches the API server and collects various metrics about the workloads, storage, and nodes in the clusters.

From the API server, the agent captures all changes to the following objects:

• Nodes

• Services and endpoints

• Pods

• PersistentVolumes

• PersistentVolumeClaims

To understand the node and workload utilization, the agent collects the following metrics exposed by Kubelet on each node:

• container_cpu_usage_seconds_total

• container_memory_rss

• container_fs_reads_total

• container_fs_writes_total

• container_fs_usage_bytes

• container_network_receive_bytes_total

• container_network_transmit_bytes_total

• kubelet_volume_stats_used_bytes

• kubelet_volume_stats_capacity_bytes

• node_cpu_seconds_total

• node_memory_MemTotal_bytes

• node_memory_MemFree_bytes

• node_memory_Buffers_bytes

• node_memory_Cached_bytes

• node_network_receive_bytes_total

• node_network_transmit_bytes_total

• node_filesystem_free_bytes

• node_filesystem_size_bytes

• node_disk_reads_completed_total

• node_disk_writes_completed_total

To accurately determine the pricing of nodes and volumes in cloud environments, the agent will (optionally) query the following APIs on AWS:

• EC2 / DescribeInstances
• EC2 / DescribeVolumes
• Cost Explorer / GetCostAndUsage

On GCP, it will use the following APIs:

• instances.aggregatedList