Google CloudDataproc MetastoreAnalytics

`google_dataproc_metastore_service` cost estimation

A managed Hive metastore for data lakes. Priced as a flat hourly service fee by tier, billed continuously while the service runs.

A google_dataproc_metastore_service provides a managed, highly-available Apache Hive metastore, the central catalog of table and schema metadata that Dataproc, BigQuery, and other engines query against a data lake. It removes the need to run and back up your own metastore database.

Cost is a flat hourly service fee by tier, billed continuously once the service exists. The Developer tier is a lower-cost, non-HA option for testing; the Enterprise tier provides high availability and scales to production metadata volumes at a higher hourly rate, on the order of $0.50/hour (~$365/month). The fee is independent of how many queries hit the metastore, it's provisioned capacity, not usage-metered.

c3x reads the tier and prices the service-hours at 730 hours/month. There's no scale-to-zero, so the tier's monthly figure is a standing cost. The optimization is matching the tier to need (Developer for non-production) and consolidating, one shared metastore can serve multiple Dataproc clusters and query engines rather than running several.

Terraform example

A minimal but realistic configuration that C3X can estimate.

resource "google_dataproc_metastore_service" "lake" {
  service_id = "lake-metastore"
  location   = "us-central1"
  tier       = "DEVELOPER"

  hive_metastore_config {
    version = "3.1.2"
  }
}

Pricing dimensions

What you actually pay for when you provision google_dataproc_metastore_service.

Dimension	Unit	What's being charged
Service hours	per hour	Flat hourly fee by tier, billed continuously. c3x reads tier and assumes 730 hours/month. ~$0.50/hour Enterprise (~$365/month); lower for Developer
Tier	Developer / Enterprise	Enterprise adds high availability and production scale; Developer is a cheaper non-HA option for testing.

Sample C3X output

Example output from c3x estimate (Enterprise tier):

google_dataproc_metastore_service.lake
└─ Service hours (Enterprise)   730  hours    $365.00

OVERALL TOTAL                                 $365.00

Optimization tips

Common ways to reduce google_dataproc_metastore_service cost without changing the workload.

Use the Developer tier for non-production

Tier delta

The Developer tier is a cheaper, non-HA metastore fine for dev, testing, and experimentation. Reserve the Enterprise tier (and its higher hourly fee) for production data lakes that need HA.

Share one metastore across clusters and engines

Per avoided metastore

A single metastore serves many Dataproc clusters, BigQuery, and other engines as a shared catalog. Don't run a separate metastore per cluster, consolidate to one and point everything at it.

Delete metastores from retired lakes

Full service fee

The hourly fee accrues continuously. A metastore left behind after a data-lake project ends quietly bills ~$365/month on Enterprise. Decommission unused services.

FAQ

How does c3x estimate Dataproc Metastore cost?

It reads the tier and prices the flat service-hour fee at 730 hours/month. The fee is provisioned (not per-query), so the estimate is the continuous monthly cost for the chosen tier.

Why does the metastore cost money even when idle?

It's a provisioned, always-on service billed by the hour, not per query. Whether or not jobs are querying metadata, the tier's hourly fee accrues continuously.

What's the difference between the tiers?

Developer is a lower-cost, non-HA metastore for testing and small workloads. Enterprise provides high availability and production scale at a higher hourly rate (~$365/month).

Can one metastore serve multiple clusters?

Yes, and it should. A single metastore acts as a shared catalog for many Dataproc clusters and query engines. Consolidating onto one avoids paying the service fee multiple times.

Is this required to run Dataproc?

No. Dataproc clusters can use a local or cluster-internal metastore. The managed Dataproc Metastore is for a persistent, shared, HA catalog across clusters, worth its fee when metadata must outlive any single cluster.