# gRPC Health Checks

This chapter describes how to configure health checks for gRPC servers in a load‑balanced upstream server group.

## Introduction

NGINX Plus can monitor the health of upstream servers by making active health checks. NGINX Plus R23 supports the gRPC health checking protocol so that upstream gRPC services can be tested for their ability to handle new requests. This is particularly important in dynamic and containerized environments. When adding a new instance of a gRPC service, it is important that requests are sent to the fully operating service.

## Prerequisites

• You have configured an upstream group of servers that handles gRPC network traffic and specified a shared memory zone that keeps the state of these servers, for example:

stream {
#...
upstream grpc_backend {
zone   grpc_backend 64k;
server 10.0.0.1:50051;
server 10.0.0.2:50051;
}
#...
}

• You have configured a server that routes gRPC requests to the upstream server group:

location /grpc {
grpc_pass grpc://grpc_backend;
}


## gRPC Servers that accept health checking protocol

If your gRPC services support the gRPC health checking protocol, specify the type=grpc parameter of the health_check directive:

location / {
grpc_pass    grpc://grpc_backend;
health_check mandatory type=grpc;
}


In this example, according to health checking protocol, the Check method of the Health service will be invoked, and the gRPC server that respond with SERVING are considered healthy.

The mandatoryparameter ensures that the health check must pass before traffic is sent on an instance, for example, when it is introduced or reloaded. Note that the type=grpc must be specified after all other health_check parameters.

If there are several gRPC services exposed on each upstream server then the most significant service can be monitored by specifying the service name with the grpc_service parameter:

location / {
grpc_pass    grpc://grpc_backend;
health_check mandatory type=grpc grpc_service=MyStatus;
}


## gRPC Servers that do not accept health checking protocol

If your gRPC services do not implement the gRPC health checking protocol, it is still possible to perform health check by sending the Check method and expecting a particular status code. This will test that the upstream server is responding to gRPC requests. Specify the gRPC error response code with the grpc_status parameter:

location / {
grpc_pass    grpc://grpc_backend;
health_check type=grpc grpc_status=12; # 12=unimplemented
}


In this case the response with the gRPC status of 12 / UNIMPLEMENTED will be considered healthy.

Note that the type=grpc parameter is not compatible with uri or match parameters of the health_check directive. The type=grpc parameter must be specified after all other directive parameters: grpc_service and grpc_status must follow type=grpc.