Java interceptors which can be used to monitor Grpc services using Prometheus.
The features of this library include two monitoring grpc interceptors, MonitoringServerInterceptor
and MonitoringClientInterceptor
. These interceptors can be attached separately to grpc servers and client stubs respectively. For each RPC, the interceptors increment the following Prometheus metrics, broken down by method type, service name, method name, and response code:
- Server
grpc_server_started_total
: Total number of RPCs started on the server.grpc_server_handled_total
: Total number of RPCs completed on the server, regardless of success or failure.grpc_server_handled_latency_seconds
: (Optional) Histogram of response latency of rpcs handled by the server, in seconds.grpc_server_msg_received_total
: Total number of stream messages received from the client.grpc_server_msg_sent_total
: Total number of stream messages sent by the server.
- Client
grpc_client_started_total
: Total number of RPCs started on the client.grpc_client_completed
: Total number of RPCs completed on the client, regardless of success or failure.grpc_client_completed_latency_seconds
: (Optional) Histogram of rpc response latency for completed rpcs, in seconds.grpc_client_msg_received_total
: Total number of stream messages received from the server.grpc_client_msg_sent_total
: Total number of stream messages sent by the client.
Note that by passing a Configuration
instance to the interceptors, it is possible to configure the following:
- Whether or not a latency histogram is recorded for RPCs.
- Which histogram buckets to use for the latency metrics.
- Which Prometheus
CollectorRegistry
the metrics get registered with. - (Optional) Which headers you want to be applied to metrics as added labels.
The server interceptors have an identical implementation in Golang, go-grpc-prometheus, brought to you by @MWitkow.
This library is made available on the dinowernli GitHub Maven repository. Once the repository is set up, the library can be included using the following artifact id:
me.dinowernli:java-grpc-prometheus:0.3.0
In order to attach the monitoring server interceptor to your gRPC server, you can do the following:
MonitoringServerInterceptor monitoringInterceptor =
MonitoringServerInterceptor.create(Configuration.cheapMetricsOnly());
grpcServer = ServerBuilder.forPort(GRPC_PORT)
.addService(ServerInterceptors.intercept(
HelloServiceGrpc.bindService(new HelloServiceImpl()), monitoringInterceptor))
.build();
In order to attach the monitoring client interceptor to your gRPC client, you can do the following:
MonitoringClientInterceptor monitoringInterceptor =
MonitoringClientInterceptor.create(Configuration.cheapMetricsOnly());
grpcStub = HelloServiceGrpc.newStub(NettyChannelBuilder.forAddress(REMOTE_HOST, GRPC_PORT)
.intercept(monitoringInterceptor)
.build());
If you want to instruct the interceptor to use a specific header of interest, for example "header-1" as a label on all produced metrics, you can do the following, which will cause the metrics to carry an extra label "header_1" whose value is filled from the header value on each request:
MonitoringServerInterceptor monitoringInterceptor =
MonitoringServerInterceptor.create(
Configuration
.cheapMetricsOnly()
.withLabelHeaders(Arrays.asList("header-1"))
);
grpcServer = ServerBuilder.forPort(GRPC_PORT)
.addService(ServerInterceptors.intercept(
HelloServiceGrpc.bindService(new HelloServiceImpl()), monitoringInterceptor))
.build();
If you're using Spring Boot 2 with micrometer-registry-prometheus you should inject the CollectorRegistry that is already provided in the application context:
@Autowired
private CollectorRegistry collectorRegistry;
// use the provided registry
MonitoringServerInterceptor monitoringInterceptor =
MonitoringServerInterceptor.create(Configuration.cheapMetricsOnly().withCollectorRegistry(collectorRegistry));