gRPC

Overview

The gRPC monitor calls the standard grpc.health.v1.Health/Check method on your service and reads back its serving status. Unlike a TCP check on the same port, this proves the whole stack works — TCP, TLS, HTTP/2 and gRPC itself — and that the service says it is serving.

Available on the Pro plan and above.

Target format

The monitor’s target is written as one string:

[scheme://]host:port[/service]

Part	Meaning
scheme	grpcs or https → TLS. grpc or http → plaintext. Omitted → TLS when the port is 443, plaintext otherwise.
host:port	Where the service listens. Without a port, 443 is used for TLS and 80 for plaintext.
/service	Optional service name passed in the health request. Leave it out for the server’s overall health, per the health-checking protocol.

Examples:

grpcs://api.example.com:443              overall server health, over TLS
grpcs://api.example.com/orders.v1.Orders one specific service
grpc://internal.svc:9090                 plaintext, in-cluster
api.example.com:443                      TLS inferred from the port

How a result is decided

Health response	Result
SERVING	Up
NOT_SERVING	Down, with the reported status in the error
SERVICE_UNKNOWN	Down — the name in /service isn’t registered on that server
Unimplemented	Up — the server answered over gRPC but exposes no health service. The transport, HTTP/2 and gRPC stack all work, which is what an uptime check asserts; the error field records it so you can see it in the event log.
Connect, TLS or timeout error	Down

The retry window and the majority rule then apply exactly as for any other monitor — see Status & uptime.

Settings

Setting	Description	Default
Target	[scheme://]host:port[/service]	—
Port	Overrides the port in the target	from the target
Interval	Check frequency	shortest your plan allows
Timeout	Fixed check budget (not configurable)	5s
Regions	Where the check runs from	one, per your plan

Use cases

• Internal APIs — A gRPC backend that has no HTTP surface to check.
• Service meshes — Confirm a specific service name is registered and serving, not just that the port accepts connections.
• TLS termination — A plaintext grpc:// target and a grpcs:// one on the same service tell you whether a problem is in the transport or the app.

Private services

A gRPC service that isn’t reachable from the internet is monitored from a private agent — the check is identical, only the agent doing it changes. Public regions refuse targets that resolve to private addresses.

Metrics

The full call duration (connect + TLS + health call) is recorded on every check and feeds the same latency chart, percentiles and alerts as an HTTP monitor.