Cloud cost
Allocatable capacity
What is left for pods to request on a Kubernetes node after the kubelet, container runtime, OS, and DaemonSets reserve their share.
Last updated
Definition
A node's advertised capacity, its full vCPU and memory, is never what you can schedule. The kube-reserved and system-reserved settings, the eviction threshold, and any DaemonSets that run on every node each take a slice first. Allocatable is capacity minus that reservation. It is noticeably smaller than the number on the instance spec sheet.
You pay for full node capacity every hour. It does not matter how much of the allocatable slice is requested or used. So model a cluster's cost with four numbers, not one: capacity, allocatable, requested, and used.
Where it shows up
Allocatable capacity appears on the node object. Run kubectl describe node. It lists a Capacity section with the full advertised vCPU and memory. Below it, an Allocatable section is already reduced by kube-reserved, system-reserved, and the eviction threshold. This happens before any application pod is scheduled onto the node.
What makes it expensive
Teams size a cluster from a node's advertised capacity, not its allocatable slice. Then they wonder why fewer pods fit than expected. Both numbers sit on the same node record, a few lines apart. The reservation between them is often larger than anyone budgeted. Size against the lower number.
Related
More in Cloud cost
- Committed use discount (CUD)
- Reserved instance (RI)
- Savings plan (SP)
- On-demand vs. spot instances
- Data egress
- Amortized, blended, and unblended cost
- Billed vs. effective vs. list vs. contracted cost
- Rightsizing
- Idle and orphaned resources
- Storage class
- Lifecycle policy
- Incremental snapshot
- Requests vs. limits
- Cluster idle
- Workload idle
- Bin packing
- Cost allocation tags
- Tag coverage
- Unallocated spend
- Showback vs. chargeback
- Unit economics
- RI/SP coverage and utilization
- Observability tax
- Metric cardinality
- High-water-mark billing
- Log ingest vs. index
- Snowflake credit
- Auto-suspend / resume minimum
- BigQuery slot
- Bytes scanned & partition pruning
- Time Travel & Fail-safe
The meter you don't watch still grows.
Waste rarely hides in the service you check every week. CostMon watches every meter on your stack, including the ones you forgot were running.