resource-calculator

Kubernetes resource metrics calculator, plus an inspect command that lists managed cloud databases with their allocated CPU and memory.

Install

Build the binary with the Makefile (it runs inside the pinned build image):

make build            # produces bin/resource-calculator-<os>-<arch>

The binary runs standalone or as a kubectl plugin: put it on your PATH named kubectl-resource_calculator and call kubectl resource-calculator ....

Commands

• calculate: sum CPU, memory and storage of workloads in a Kubernetes cluster.
• convert: convert KubeDB v1alpha1 resources to v1alpha2.
• check-deprecated: list installed KubeDB resources on deprecated versions.
• inspect: list managed cloud databases with their allocated CPU and memory, plus inspect kubedb (per-object listing of KubeDB-managed cluster resources with memory limits) and inspect operators (self-hosted scan) -- see the user guide below.

Architecture and design notes live in DESIGN.md.

Kubernetes cluster commands

These operate on the cluster in your current kubeconfig context (add --all to sweep every context):

# sum CPU / memory / storage by kind (-o text | json | yaml)
resource-calculator calculate -o text

# list each KubeDB-managed cluster resource with its memory limit, one row per object
resource-calculator inspect kubedb -o text

# list KubeDB resources still on the v1alpha1 API
resource-calculator check-deprecated

# convert KubeDB v1alpha1 resources to v1alpha2 YAML on disk
resource-calculator convert --dir ./converted

inspect kubedb walks the same set of GVKs as calculate (every kind registered in kmodules.xyz/resource-metrics, picking the highest available API version per GroupKind) and prints one row per object with its group/kind, namespace, name, UID, age and memory limit. Filter by API group with --apiGroups, sweep every kubeconfig context with --all, and switch to JSON or YAML with the inherited -o/--output flag.

inspect: list managed databases with allocated CPU and memory

inspect inventories the managed databases on a public cloud account or a DBaaS organization and lists each one with its allocated CPU and memory, plus the estate totals.

Every managed database is reduced to its allocated CPU and memory per node times its node count (counted as replicas x size per replica). A 3 replica PostgreSQL with 8 GiB per replica counts as 24 GiB. inspect lists each discovered database with that allocation and sums the whole estate.

Supported providers: aws, azure, gcp, oci, atlas (MongoDB Atlas), elastic (Elastic Cloud), clickhouse (ClickHouse Cloud), and all.

Quick start

Live, using the provider's official SDK and your existing cloud credentials:

# AWS: default credential chain, every enabled region
resource-calculator inspect aws --all-regions

# GCP: a specific project
resource-calculator inspect gcp --account=my-project

# MongoDB Atlas: service account
resource-calculator inspect atlas \
  --atlas-client-id=$ATLAS_CLIENT_ID --atlas-client-secret=$ATLAS_CLIENT_SECRET

Offline, from exported JSON (no credentials, handy for CI or sharing):

resource-calculator inspect aws --source=file --from-file=aws-bundle.json -o json

How databases are discovered

--source selects the mechanism (priority order sdk, cli, rest, file):

• sdk (used by auto): the vendor's official Go SDK, built in for AWS, Azure, GCP, OCI, Atlas and Elastic. ClickHouse Cloud uses REST (no official Go SDK).
• cli: shell out to your installed aws/az/gcloud/oci CLI.
• rest: call the vendor REST API directly (Atlas, Elastic, ClickHouse).
• file: parse exported JSON with --from-file. No live calls or credentials.

auto (the default) uses --from-file when given, otherwise the SDK (REST for ClickHouse). Credentials per provider and the --from-file bundle formats are documented in docs/inspect.md.

Self-hosted operators

inspect operators scans the current cluster for self-hosted databases and lists their allocated CPU and memory. It finds databases run by alternative (non-KubeDB) operators (CloudNativePG, Zalando, Percona, Strimzi, ECK, Altinity, and more), detected by their CRDs, and databases deployed from Bitnami, Chainguard or Docker Hardened Images, detected by container image. Both use the controller-runtime client and unstructured objects, with no dependency on those projects:

resource-calculator inspect operators

See docs/inspect.md for the full operator list and behavior.

Output

The text output is a table with one row per database (columns include provider, service, engine, name, region, node type, CPU and memory per node, node count, and the per-database totals), ending with a totals line that reports the count of databases and nodes and the total vCPU and memory allocated across the estate. For cloud providers an estimated managed monthly cost is shown as an informational column: a memory-normalized list-price estimate, labelled as an estimate. Use -o json or -o yaml for the full report.

Useful flags

flag	meaning
--account	AWS profile / Azure subscription / GCP project / OCI compartment OCID
--all-regions	scan every enabled region (AWS)
--org	organization / all-accounts scan where supported (AWS)
--count-standby	count HA standbys / Multi-AZ mirrors as nodes (default true)
--include-non-data	include non-data nodes (for example OpenSearch dedicated masters)
-o, --output	text (default), json, or yaml

Totals show the database and node counts and the total vCPU and memory allocated. Full reference, including per-provider credentials, the JSON bundle formats, and node-counting rules: docs/inspect.md.