jina.serve.runtimes.head package#

Module contents#

class jina.serve.runtimes.head.HeadRuntime(args, **kwargs)[source]#

Bases: AsyncNewLoopRuntime, ABC

Runtime is used in head pods. It responds to Gateway requests and sends to uses_before/uses_after and its workers

Initialize grpc server for the head runtime. :type args: Namespace :param args: args from CLI :param kwargs: keyword args

DEFAULT_POLLING = 1#
async async_setup()[source]#

Wait for the GRPC server to start

async async_run_forever()[source]#

Block until the GRPC server is terminated

async async_cancel()[source]#

Stop the GRPC server

async async_teardown()[source]#

Close the connection pool

async process_single_data(request, context)[source]#

Process the received requests and return the result as a new request

Parameters:

  • request (DataRequest) – the data request to process

  • context – grpc context

Return type:

DataRequest

Returns:

the response request

async process_data(requests, context)[source]#

Process the received data request and return the result as a new request

Parameters:

  • requests (List[DataRequest]) – the data requests to process

  • context – grpc context

Return type:

DataRequest

Returns:

the response request

async endpoint_discovery(empty, context)[source]#

Uses the connection pool to send a discover endpoint call to the workers

Parameters:

  • empty – The service expects an empty protobuf message

  • context – grpc context

Return type:

EndpointsProto

Returns:

the response request

static activate(**kwargs)#

Activate the runtime, does not apply to these runtimes

Parameters:

kwargs – extra keyword arguments

static is_ready(ctrl_address, **kwargs)#

Check if status is ready.

Parameters:

  • ctrl_address (str) – the address where the control request needs to be sent

  • kwargs – extra keyword arguments

Return type:

bool

Returns:

True if status is ready else False.

run_forever()#

Running method to block the main thread.

Run the event loop until a Future is done.

teardown()#

Call async_teardown() and stop and close the event loop.

classmethod wait_for_ready_or_shutdown(timeout, ready_or_shutdown_event, ctrl_address, **kwargs)#

Check if the runtime has successfully started

Parameters:

  • timeout (Optional[float]) – The time to wait before readiness or failure is determined

  • ctrl_address (str) – the address where the control message needs to be sent

  • ready_or_shutdown_event (Union[Event, Event]) – the multiprocessing event to detect if the process failed or is ready

  • kwargs – extra keyword arguments

Returns:

True if is ready or it needs to be shutdown