io.servicetalk.grpc.health

Class DefaultHealthService

Object

DefaultHealthService

All Implemented Interfaces:

AsyncCloseable, GrpcBindableService<Health.HealthService>, GrpcService, Health.CheckRpc, Health.HealthService, Health.WatchRpc

public final class DefaultHealthService
extends Object
implements Health.HealthService

Implementation of Health.HealthService which targets gRPC health checking that provides accessors to set/clear status for arbitrary services.

Field Summary

Fields

Modifier and Type	Field and Description
static String	OVERALL_SERVICE_NAME The name of the service corresponding to the overall health status.

Fields inherited from interface Health.CheckRpc

PATH

Fields inherited from interface Health.WatchRpc

PATH

Constructor Summary

Constructors

Constructor and Description
DefaultHealthService() Create a new instance.
DefaultHealthService(Predicate<String> watchAllowed) Create a new instance.

Method Summary

Modifier and Type	Method and Description
Single<HealthCheckResponse>	check(GrpcServiceContext ctx, HealthCheckRequest request) If the requested service is unknown, the call will fail with status NOT_FOUND.
boolean	clearStatus(String service) Clears the health status record of a service.
boolean	setStatus(String service, HealthCheckResponse.ServingStatus status) Updates the status of the server.
boolean	terminate() All services will be marked as HealthCheckResponse.ServingStatus.NOT_SERVING, and future updates to services will be prohibited.
Publisher<HealthCheckResponse>	watch(GrpcServiceContext ctx, HealthCheckRequest request) Performs a watch for the serving status of the requested service.

Methods inherited from class Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Methods inherited from interface Health.HealthService

bindService, methodDescriptors

Methods inherited from interface Health.CheckRpc

methodDescriptor

Methods inherited from interface Health.WatchRpc

methodDescriptor

Methods inherited from interface GrpcService

closeAsync

Methods inherited from interface AsyncCloseable

closeAsyncGracefully

Field Detail

OVERALL_SERVICE_NAME

public static final String OVERALL_SERVICE_NAME

The name of the service corresponding to the overall health status.

See Also:

Constant Field Values

Constructor Detail

DefaultHealthService

public DefaultHealthService()

Create a new instance. Service OVERALL_SERVICE_NAME state is set to HealthCheckResponse.ServingStatus.SERVING.

DefaultHealthService

public DefaultHealthService(Predicate<String> watchAllowed)

Create a new instance. Service OVERALL_SERVICE_NAME state is set to HealthCheckResponse.ServingStatus.SERVING.

Parameters:

watchAllowed - Predicate that determines if a watch(GrpcServiceContext, HealthCheckRequest) request that doesn't match an existing service will succeed or fail with GrpcStatusCode.FAILED_PRECONDITION. This can be used to bound memory by restricting watches to expected service names.

Method Detail

check

public Single<HealthCheckResponse> check(GrpcServiceContext ctx,
                                         HealthCheckRequest request)

Description copied from interface: Health.CheckRpc

  If the requested service is unknown, the call will fail with status
  NOT_FOUND.
 

Specified by:

check in interface Health.CheckRpc

Parameters:

ctx - context associated with this service and request.

request - the request from the client.

Returns:

a Single which sends the response to the client when it terminates.

watch

public Publisher<HealthCheckResponse> watch(GrpcServiceContext ctx,
                                            HealthCheckRequest request)

Description copied from interface: Health.WatchRpc

  Performs a watch for the serving status of the requested service.
  The server will immediately send back a message indicating the current
  serving status.  It will then subsequently send a new message whenever
  the service's serving status changes.

  If the requested service is unknown when the call is received, the
  server will send a message setting the serving status to
  SERVICE_UNKNOWN but will *not* terminate the call.  If at some
  future point, the serving status of the service becomes known, the
  server will send a new message with the service's serving status.

  If the call terminates with status UNIMPLEMENTED, then clients
  should assume this method is not supported and should not retry the
  call.  If the call terminates with any other status (including OK),
  clients should retry the call with appropriate exponential backoff.
 

Specified by:

watch in interface Health.WatchRpc

Parameters:

ctx - context associated with this service and request.

request - the request from the client.

Returns:

used to write a stream of type HealthCheckResponse to the client.

setStatus

public boolean setStatus(String service,
                         HealthCheckResponse.ServingStatus status)

Updates the status of the server.

Parameters:

service - the name of some aspect of the server that is associated with a health status. This name can have no relation with the gRPC services that the server is running with. It can also be an empty String "" per the gRPC specification.

status - is one of the values HealthCheckResponse.ServingStatus.SERVING, HealthCheckResponse.ServingStatus.NOT_SERVING, and HealthCheckResponse.ServingStatus.UNKNOWN.

Returns:

true if this change was applied. false if it was not due to terminate().

clearStatus

public boolean clearStatus(String service)

Clears the health status record of a service. The health service will respond with NOT_FOUND error on checking the status of a cleared service.

Parameters:

service - the name of some aspect of the server that is associated with a health status. This name can have no relation with the gRPC services that the server is running with. It can also be an empty String "" per the gRPC specification.

Returns:

true if this call removed a service. false if service wasn't found.

terminate

public boolean terminate()

All services will be marked as HealthCheckResponse.ServingStatus.NOT_SERVING, and future updates to services will be prohibited. This method is meant to be called prior to server shutdown as a way to indicate that clients should redirect their traffic elsewhere.

Returns:

true if this call terminated this service. false if it was not due to previous call to this method.