mirror of
https://github.com/wassname/ray.git
synced 2026-07-19 11:27:32 +08:00
implementing worker.py and miscellaneous changes
This commit is contained in:
+5
-1
@@ -2,15 +2,19 @@ cmake_minimum_required(VERSION 2.8)
|
||||
|
||||
project(orchestra)
|
||||
|
||||
list(APPEND CMAKE_MODULE_PATH ${PROJECT_SOURCE_DIR}/cmake/Modules)
|
||||
|
||||
find_package(Protobuf REQUIRED)
|
||||
find_package(PythonInterp REQUIRED)
|
||||
find_package(PythonLibs REQUIRED)
|
||||
find_package(NumPy REQUIRED)
|
||||
|
||||
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=c++11")
|
||||
|
||||
include_directories("${CMAKE_SOURCE_DIR}/include")
|
||||
include_directories("/usr/local/include")
|
||||
include_directories("${PYTHON_INCLUDE_DIRS}")
|
||||
include_directories("/home/pcmoritz/anaconda/pkgs/numpy-1.9.2-py27_0/lib/python2.7/site-packages/numpy/core/include/")
|
||||
include_directories("${NUMPY_INCLUDE_DIR}")
|
||||
|
||||
set(PROTO_PATH "${CMAKE_SOURCE_DIR}/protos")
|
||||
|
||||
|
||||
@@ -41,5 +41,10 @@ def start_objstore(host, port):
|
||||
all_processes.append((p, port))
|
||||
|
||||
def start_worker(test_path, host, scheduler_port, worker_port, objstore_port):
|
||||
p = subprocess.Popen(["python", test_path, host, str(scheduler_port), str(worker_port), str(objstore_port)])
|
||||
p = subprocess.Popen(["python",
|
||||
test_path,
|
||||
"--ip_address=" + host,
|
||||
"--scheduler_port=" + str(scheduler_port),
|
||||
"--objstore_port=" + str(objstore_port),
|
||||
"--worker_port=" + str(worker_port)])
|
||||
all_processes.append((p, worker_port))
|
||||
|
||||
+110
-13
@@ -1,25 +1,122 @@
|
||||
import typing
|
||||
|
||||
import orchpy
|
||||
|
||||
class Worker(object):
|
||||
"""The methods in this class are considered unexposed to the user. The functions outside of this class are considered exposed."""
|
||||
|
||||
def __init__(self):
|
||||
self.functions = {}
|
||||
self.connected = False
|
||||
self.handle = None
|
||||
|
||||
def connect(self, scheduler_addr, objstore_addr, worker_addr):
|
||||
self.worker = orchpy.lib.create_worker(scheduler_addr, objstore_addr, worker_addr)
|
||||
def put_object(self, objref, value):
|
||||
"""Put `value` in the local object store with objref `objref`. This assumes that the value for `objref` has not yet been placed in the local object store."""
|
||||
object_capsule = orchpy.lib.serialize_object(value)
|
||||
orchpy.lib.put_object(self.handle, objref, object_capsule)
|
||||
|
||||
def call(self, name, args):
|
||||
return orchpy.lib.remote_call(self.worker, name, args)
|
||||
def get_object(self, objref):
|
||||
"""Return the value from the local object store for objref `objref`. This will block until the value for `objref` has been written to the local object store."""
|
||||
object_capsule = orchpy.lib.get_object(self.handle, objref)
|
||||
return orchpy.lib.deserialize_object(object_capsule)
|
||||
|
||||
def pull(self, objref):
|
||||
pass
|
||||
def register_function(self, function):
|
||||
"""Notify the scheduler that this worker can execute the function with name `func_name`. Store the function `function` locally."""
|
||||
orchpy.lib.register_function(self.handle, function.func_name, len(function.return_types))
|
||||
self.functions[function.func_name] = function
|
||||
|
||||
def push(self, objref):
|
||||
pass
|
||||
def remote_call(self, func_name, args):
|
||||
"""Tell the scheduler to schedule the execution of the function with name `func_name` with arguments `args`. Retrieve object references for the outputs of the function from the scheduler and immediately return them."""
|
||||
call_capsule = orchpy.lib.serialize_call(func_name, args)
|
||||
return orchpy.lib.remote_call(self.handle, call_capsule)
|
||||
|
||||
def main_loop(self):
|
||||
while True:
|
||||
call = self.worker.wait_for_next_task()
|
||||
name, args, results = orchpy.lib.deserialize_call(call)
|
||||
# We make `global_worker` a global variable so that there is one worker per worker process.
|
||||
global_worker = Worker()
|
||||
|
||||
# TODO: finish this
|
||||
def connect(scheduler_addr, objstore_addr, worker_addr, worker=global_worker):
|
||||
if worker.connected:
|
||||
raise Exception("Worker called connect, but worker is already connected")
|
||||
worker.handle = orchpy.lib.create_worker(scheduler_addr, objstore_addr, worker_addr)
|
||||
worker.connected = True
|
||||
|
||||
def pull(objref, worker=global_worker):
|
||||
object_capsule = orchpy.lib.pull_object(worker.handle, objref)
|
||||
return orchpy.lib.deserialize_object(object_capsule)
|
||||
|
||||
def push(value, worker=global_worker):
|
||||
object_capsule = orchpy.lib.serialize_object(value)
|
||||
return orchpy.lib.push_object(worker.handle, object_capsule)
|
||||
|
||||
def main_loop(worker=global_worker):
|
||||
if not worker.connected:
|
||||
raise Exception("Worker is attempting to enter main_loop but has not been connected yet.")
|
||||
orchpy.lib.start_worker_service(worker.handle)
|
||||
while True:
|
||||
call = orchpy.lib.wait_for_next_task(worker.handle)
|
||||
func_name, args, return_objrefs = orchpy.lib.deserialize_call(call)
|
||||
arguments = get_arguments_for_execution(worker.functions[func_name], args, worker) # get args from objstore
|
||||
outputs = worker.functions[func_name].executor(arguments) # execute the function
|
||||
store_outputs_in_objstore(return_objrefs, outputs, worker) # store output in local object store
|
||||
# TODO(rkn): notify the scheduler that the task has completed, orchpy.lib.notify_task_completed(worker.handle)
|
||||
|
||||
def distributed(arg_types, return_types, worker=global_worker):
|
||||
def distributed_decorator(func):
|
||||
def func_executor(arguments):
|
||||
"""This is what gets executed remotely on a worker after a distributed function is scheduled by the scheduler."""
|
||||
print "Calling function {} with arguments {}".format(func.__name__, arguments)
|
||||
result = func(*arguments)
|
||||
if len(return_types) != 1 and len(result) != len(return_types):
|
||||
raise Exception("The @distributed decorator for function {} has {} return values with types {}, but {} returned {} values.".format(func.__name__, len(return_types), return_types, func.__name__, len(result)))
|
||||
return result
|
||||
def func_call(*args):
|
||||
"""This is what gets run immediately when a worker calls a distributed function."""
|
||||
# TODO(rkn): check types
|
||||
return worker.remote_call(func_call.func_name, list(args))
|
||||
func_call.func_name = "{}.{}".format(func.__module__, func.__name__)
|
||||
func_call.executor = func_executor
|
||||
func_call.arg_types = arg_types
|
||||
func_call.return_types = return_types
|
||||
return func_call
|
||||
return distributed_decorator
|
||||
|
||||
# helper method, this should not be called by the user
|
||||
def get_arguments_for_execution(function, args, worker=global_worker):
|
||||
arguments = []
|
||||
# check the number of args
|
||||
if len(args) != len(function.types) and function.types[-1] is not None:
|
||||
raise Exception("Function {} expects {} arguments, but received {}.".format(function.__name__, len(function.types), len(args)))
|
||||
elif len(args) < len(function.types) - 1 and function.types[-1] is None:
|
||||
raise Exception("Function {} expects at least {} arguments, but received {}.".format(function.__name__, len(function.types) - 1, len(args)))
|
||||
|
||||
for (i, arg) in enumerate(args):
|
||||
print "Pulling argument {} for function {}.".format(i, function.__name__)
|
||||
if i < len(function.types) - 1:
|
||||
expected_type = function.types[i]
|
||||
elif i == len(function.types) - 1 and function.types[-1] is not None:
|
||||
expected_type = function.types[-1]
|
||||
elif function.types[-1] is None and len(function.types > 1):
|
||||
expected_type = function.types[-2]
|
||||
else:
|
||||
assert False, "This code should be unreachable."
|
||||
|
||||
argument = worker.get_object(arg) if type(arg) == orchpy.ObjRef else arg
|
||||
if type(arg) == orchpy.ObjRef:
|
||||
# get the object from the local object store
|
||||
# TODO(rkn): Do we know that it is already there? Maybe we should call pull(arg, worker).
|
||||
argument = worker.get_object(arg)
|
||||
else:
|
||||
# pass the argument by value
|
||||
argument = arg
|
||||
|
||||
if expected_type != type(argument):
|
||||
raise Exception("Argument {} for function {} has type {} but an argument of type {} was expected.".format(i, function.__name__, type(argument), arg_type))
|
||||
arguments.append(argument)
|
||||
return arguments
|
||||
|
||||
# helper method, this should not be called by the user
|
||||
def store_outputs_in_objstore(objrefs, outputs, worker=global_worker):
|
||||
if len(objrefs) == 1:
|
||||
worker.put_object(objrefs[0], outputs)
|
||||
else:
|
||||
for i in range(len(objrefs)):
|
||||
worker.put_object(objrefs[i], outputs[i])
|
||||
|
||||
@@ -38,7 +38,8 @@ message RemoteCallReply {
|
||||
}
|
||||
|
||||
message PullObjRequest {
|
||||
uint64 objref = 1;
|
||||
uint64 workerid = 1;
|
||||
uint64 objref = 2;
|
||||
}
|
||||
|
||||
message PushObjRequest {
|
||||
|
||||
+135
-20
@@ -10,13 +10,20 @@
|
||||
#include "types.pb.h"
|
||||
#include "worker.h"
|
||||
|
||||
extern "C" {
|
||||
// Error handling
|
||||
|
||||
static PyObject *OrchPyError;
|
||||
}
|
||||
|
||||
// extracts a pointer from a python C API capsule
|
||||
template<typename T>
|
||||
T* get_pointer_or_fail(PyObject* capsule, const char* name) {
|
||||
if (PyCapsule_IsValid(capsule, name)) {
|
||||
T* result = static_cast<T*>(PyCapsule_GetPointer(capsule, name));
|
||||
return static_cast<T*>(PyCapsule_GetPointer(capsule, name));
|
||||
} else {
|
||||
std::cout << "not a valid capsule" << std::endl;
|
||||
PyErr_SetString(OrchPyError, "not a vaid capsule");
|
||||
return NULL;
|
||||
}
|
||||
}
|
||||
|
||||
@@ -97,15 +104,12 @@ static PyTypeObject PyObjRefType = {
|
||||
|
||||
// create PyObjRef from C++ (could be made more efficient if neccessary)
|
||||
PyObject* make_pyobjref(ObjRef objref) {
|
||||
PyObject* result = PyObjRef_new(&PyObjRefType, NULL, NULL);
|
||||
PyObjRef_init((PyObjRef*) result, Py_BuildValue("i", objref), NULL);
|
||||
PyObject* arglist = Py_BuildValue("(i)", objref);
|
||||
PyObject* result = PyObject_CallObject((PyObject*) &PyObjRefType, arglist);
|
||||
Py_DECREF(arglist);
|
||||
return result;
|
||||
}
|
||||
|
||||
// Error handling
|
||||
|
||||
static PyObject *OrchPyError;
|
||||
|
||||
// Serialization
|
||||
|
||||
// serialize will serialize the python object val into the protocol buffer
|
||||
@@ -123,8 +127,9 @@ int serialize(PyObject* val, Obj* obj) {
|
||||
List* data = obj->mutable_list_data();
|
||||
for (size_t i = 0, size = PyList_Size(val); i < size; ++i) {
|
||||
Obj* elem = data->add_elem();
|
||||
if (serialize(PyList_GetItem(val, i), elem) != 0)
|
||||
if (serialize(PyList_GetItem(val, i), elem) != 0) {
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
} else if (PyString_Check(val)) {
|
||||
char* buffer;
|
||||
@@ -189,17 +194,26 @@ PyObject* deserialize(const Obj& obj) {
|
||||
|
||||
PyObject* serialize_object(PyObject* self, PyObject* args) {
|
||||
Obj* obj = new Obj(); // TODO: to be freed in capsul destructor
|
||||
PyObject* val = PyTuple_GetItem(args, 0);
|
||||
if (serialize(val, obj) != 0) {
|
||||
PyObject* pyval;
|
||||
if (!PyArg_ParseTuple(args, "O", &pyval)) {
|
||||
return NULL;
|
||||
}
|
||||
if (serialize(pyval, obj) != 0) {
|
||||
PyErr_SetString(OrchPyError, "serialization: type not know"); // TODO: put a more expressive error message here
|
||||
return NULL;
|
||||
}
|
||||
return PyCapsule_New(static_cast<void*>(obj), NULL, NULL);
|
||||
return PyCapsule_New(static_cast<void*>(obj), "obj", NULL);
|
||||
}
|
||||
|
||||
PyObject* deserialize_object(PyObject* self, PyObject* args) {
|
||||
PyObject* capsule = PyTuple_GetItem(args, 0);
|
||||
Obj* obj = get_pointer_or_fail<Obj>(capsule, NULL);
|
||||
PyObject* capsule;
|
||||
if (!PyArg_ParseTuple(args, "O", &capsule)) {
|
||||
return NULL;
|
||||
}
|
||||
Obj* obj = get_pointer_or_fail<Obj>(capsule, "obj");
|
||||
if (!obj) {
|
||||
return NULL;
|
||||
}
|
||||
return deserialize(*obj);
|
||||
}
|
||||
|
||||
@@ -218,7 +232,8 @@ PyObject* serialize_call(PyObject* self, PyObject* args) {
|
||||
serialize(PyList_GetItem(arguments, i), arg);
|
||||
}
|
||||
} else {
|
||||
std::cout << "second argument is not a list" << std::endl;
|
||||
PyErr_SetString(OrchPyError, "serialize_call: second argument needs to be a list");
|
||||
return NULL;
|
||||
}
|
||||
return PyCapsule_New(static_cast<void*>(call), "call", NULL);
|
||||
}
|
||||
@@ -226,6 +241,9 @@ PyObject* serialize_call(PyObject* self, PyObject* args) {
|
||||
PyObject* deserialize_call(PyObject* self, PyObject* args) {
|
||||
PyObject* capsule = PyTuple_GetItem(args, 0);
|
||||
Call* call = get_pointer_or_fail<Call>(capsule, "call");
|
||||
if (!call) {
|
||||
return NULL;
|
||||
}
|
||||
PyObject* string = PyString_FromStringAndSize(call->name().c_str(), call->name().size());
|
||||
int argsize = call->arg_size();
|
||||
PyObject* arglist = PyList_New(argsize);
|
||||
@@ -264,6 +282,9 @@ PyObject* create_worker(PyObject* self, PyObject* args) {
|
||||
PyObject* wait_for_next_task(PyObject* self, PyObject* args) {
|
||||
PyObject* capsule = PyTuple_GetItem(args, 0);
|
||||
Worker* worker = get_pointer_or_fail<Worker>(capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
Call* call = worker->receive_next_task();
|
||||
return PyCapsule_New(static_cast<void*>(call), "call", NULL); // TODO: how is destruction going to be handled here?
|
||||
}
|
||||
@@ -275,7 +296,13 @@ PyObject* remote_call(PyObject* self, PyObject* args) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
Call* call = get_pointer_or_fail<Call>(call_capsule, "call");
|
||||
if (!call) {
|
||||
return NULL;
|
||||
}
|
||||
RemoteCallRequest request;
|
||||
request.set_allocated_call(call);
|
||||
RemoteCallReply reply = worker->remote_call(&request);
|
||||
@@ -296,18 +323,100 @@ PyObject* register_function(PyObject* self, PyObject* args) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
worker->register_function(std::string(function_name), num_return_vals);
|
||||
return worker_capsule;
|
||||
Py_RETURN_NONE;
|
||||
}
|
||||
|
||||
PyObject* pull_object(PyObject* self, PyObject* args) {
|
||||
// TODO: test this
|
||||
PyObject* push_object(PyObject* self, PyObject* args) {
|
||||
PyObject* worker_capsule;
|
||||
PyObject* objref;
|
||||
if (!PyArg_ParseTuple(args, "OO", &worker_capsule, &objref)) {
|
||||
PyObject* obj_capsule;
|
||||
if (!PyArg_ParseTuple(args, "OO", &worker_capsule, &obj_capsule)) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
return worker_capsule;
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
Obj* obj = get_pointer_or_fail<Obj>(obj_capsule, "obj");
|
||||
if (!obj) {
|
||||
return NULL;
|
||||
}
|
||||
ObjRef objref = worker->push_object(obj);
|
||||
return make_pyobjref(objref);
|
||||
}
|
||||
|
||||
// TODO: test this
|
||||
PyObject* put_object(PyObject* self, PyObject* args) {
|
||||
PyObject* worker_capsule;
|
||||
PyObject* pyobjref;
|
||||
PyObject* obj_capsule;
|
||||
if (!PyArg_ParseTuple(args, "OOO", &worker_capsule, &pyobjref, &obj_capsule)) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
Obj* obj = get_pointer_or_fail<Obj>(obj_capsule, "obj");
|
||||
if (!obj) {
|
||||
return NULL;
|
||||
}
|
||||
ObjRef objref = ((PyObjRef*) pyobjref)->val;
|
||||
worker->put_object(objref, obj);
|
||||
Py_RETURN_NONE;
|
||||
}
|
||||
|
||||
PyObject* get_object(PyObject* self, PyObject* args) {
|
||||
PyObject* worker_capsule;
|
||||
PyObject* pyobjref;
|
||||
if (!PyArg_ParseTuple(args, "OO", &worker_capsule, &pyobjref)) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
ObjRef objref = ((PyObjRef*) pyobjref)->val;
|
||||
slice s = worker->get_object(objref);
|
||||
Obj* obj = new Obj(); // TODO: Make sure this will get deleted
|
||||
obj->ParseFromString(std::string(s.data, s.len));
|
||||
return PyCapsule_New(static_cast<void*>(obj), "obj", NULL);
|
||||
}
|
||||
|
||||
// TODO: implement this
|
||||
PyObject* pull_object(PyObject* self, PyObject* args) {
|
||||
PyObject* worker_capsule;
|
||||
PyObject* pyobjref;
|
||||
if (!PyArg_ParseTuple(args, "OO", &worker_capsule, &pyobjref)) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
ObjRef objref = ((PyObjRef*) pyobjref)->val;
|
||||
slice s = worker->get_object(objref);
|
||||
Obj* obj = new Obj(); // TODO: Make sure this will get deleted
|
||||
obj->ParseFromString(std::string(s.data, s.len));
|
||||
return PyCapsule_New(static_cast<void*>(obj), "obj", NULL);
|
||||
}
|
||||
|
||||
// TODO: test this
|
||||
PyObject* start_worker_service(PyObject* self, PyObject* args) {
|
||||
PyObject* worker_capsule;
|
||||
if (!PyArg_ParseTuple(args, "O", &worker_capsule)) {
|
||||
return NULL;
|
||||
}
|
||||
Worker* worker = get_pointer_or_fail<Worker>(worker_capsule, "worker");
|
||||
if (!worker) {
|
||||
return NULL;
|
||||
}
|
||||
worker->start_worker_service();
|
||||
Py_RETURN_NONE;
|
||||
}
|
||||
|
||||
static PyMethodDef SymphonyMethods[] = {
|
||||
@@ -317,7 +426,13 @@ static PyMethodDef SymphonyMethods[] = {
|
||||
{ "deserialize_call", deserialize_call, METH_VARARGS, "deserialize a call from protocol buffers" },
|
||||
{ "create_worker", create_worker, METH_VARARGS, "connect to the scheduler and the object store" },
|
||||
{ "register_function", register_function, METH_VARARGS, "register a function with the scheduler" },
|
||||
{ "put_object", put_object, METH_VARARGS, "put a protocol buffer object (given as a capsule) on the local object store" },
|
||||
{ "get_object", get_object, METH_VARARGS, "get protocol buffer object from the local object store" },
|
||||
{ "push_object", push_object, METH_VARARGS, "push a protocol buffer object (given as a capsule) to the object store" },
|
||||
{ "pull_object" , pull_object, METH_VARARGS, "pull object with a given object id from the object store" },
|
||||
{ "wait_for_next_task", wait_for_next_task, METH_VARARGS, "get next task from scheduler (blocking)" },
|
||||
{ "remote_call", remote_call, METH_VARARGS, "call a remote function" },
|
||||
{ "start_worker_service", start_worker_service, METH_VARARGS, "start the worker service" },
|
||||
{ NULL, NULL, 0, NULL }
|
||||
};
|
||||
|
||||
|
||||
+22
-2
@@ -1,3 +1,5 @@
|
||||
#include <random>
|
||||
|
||||
#include "scheduler.h"
|
||||
|
||||
Status SchedulerService::RemoteCall(ServerContext* context, const RemoteCallRequest* request, RemoteCallReply* reply) {
|
||||
@@ -30,6 +32,24 @@ Status SchedulerService::PushObj(ServerContext* context, const PushObjRequest* r
|
||||
}
|
||||
|
||||
Status SchedulerService::PullObj(ServerContext* context, const PullObjRequest* request, AckReply* reply) {
|
||||
std::lock_guard<std::mutex> objtable_lock(objtable_lock_);
|
||||
ObjRef objref = request->objref();
|
||||
if (objref >= objtable_.size() || objtable_[objref].size() == 0) {
|
||||
std::cout << "internal error: no object with objref exists" << std::endl;
|
||||
std::exit(1);
|
||||
}
|
||||
std::mt19937 rng;
|
||||
std::uniform_int_distribution<int> uni(0, objtable_[objref].size()-1);
|
||||
ObjStoreId objstoreid = uni(rng);
|
||||
std::lock_guard<std::mutex> objstore_lock(objstores_lock_);
|
||||
|
||||
DeliverObjRequest deliver_request;
|
||||
ObjStoreId id = get_store(request->workerid());
|
||||
deliver_request.set_objstore_address(objstores_[id].address);
|
||||
deliver_request.set_objref(objref);
|
||||
AckReply deliver_reply;
|
||||
ClientContext deliver_context;
|
||||
objstores_[objstoreid].objstore_stub->DeliverObj(&deliver_context, deliver_request, &deliver_reply);
|
||||
return Status::OK;
|
||||
}
|
||||
|
||||
@@ -117,8 +137,8 @@ WorkerId SchedulerService::register_worker(const std::string& worker_address, co
|
||||
ObjStoreId objstoreid = std::numeric_limits<size_t>::max();
|
||||
objstores_lock_.lock();
|
||||
for (size_t i = 0; i < objstores_.size(); ++i) {
|
||||
std::cout << "adress: " << objstores_[i].address << std::endl;
|
||||
std::cout << "my adress: " << objstore_address << std::endl;
|
||||
std::cout << "address: " << objstores_[i].address << std::endl;
|
||||
std::cout << "my address: " << objstore_address << std::endl;
|
||||
if (objstores_[i].address == objstore_address) {
|
||||
objstoreid = i;
|
||||
}
|
||||
|
||||
+15
-4
@@ -37,7 +37,17 @@ void Worker::register_worker(const std::string& worker_address, const std::strin
|
||||
return;
|
||||
}
|
||||
|
||||
ObjRef Worker::push_obj(const Obj* obj) {
|
||||
slice Worker::pull_object(ObjRef objref) {
|
||||
PullObjRequest request;
|
||||
request.set_workerid(workerid_);
|
||||
request.set_objref(objref);
|
||||
AckReply reply;
|
||||
ClientContext context;
|
||||
Status status = scheduler_stub_->PullObj(&context, request, &reply);
|
||||
return get_object(objref);
|
||||
}
|
||||
|
||||
ObjRef Worker::push_object(const Obj* obj) {
|
||||
// first get objref for the new object
|
||||
PushObjRequest push_request;
|
||||
PushObjReply push_reply;
|
||||
@@ -45,11 +55,11 @@ ObjRef Worker::push_obj(const Obj* obj) {
|
||||
Status push_status = scheduler_stub_->PushObj(&push_context, push_request, &push_reply);
|
||||
ObjRef objref = push_reply.objref();
|
||||
// then stream the object to the object store
|
||||
put_obj(objref, obj);
|
||||
put_object(objref, obj);
|
||||
return objref;
|
||||
}
|
||||
|
||||
slice Worker::get_serialized_obj(ObjRef objref) {
|
||||
slice Worker::get_object(ObjRef objref) {
|
||||
ClientContext context;
|
||||
GetObjRequest request;
|
||||
request.set_objref(objref);
|
||||
@@ -62,7 +72,8 @@ slice Worker::get_serialized_obj(ObjRef objref) {
|
||||
return slice;
|
||||
}
|
||||
|
||||
void Worker::put_obj(ObjRef objref, const Obj* obj) {
|
||||
// TODO: Do this with shared memory
|
||||
void Worker::put_object(ObjRef objref, const Obj* obj) {
|
||||
ObjChunk chunk;
|
||||
std::string data;
|
||||
obj->SerializeToString(&data);
|
||||
|
||||
+6
-4
@@ -48,11 +48,13 @@ class Worker {
|
||||
// send request to the scheduler to register this worker
|
||||
void register_worker(const std::string& worker_address, const std::string& objstore_address);
|
||||
// push object to local object store, register it with the server and return object reference
|
||||
ObjRef push_obj(const Obj* obj);
|
||||
// retrieve serialized object from local object store
|
||||
slice get_serialized_obj(ObjRef objref);
|
||||
ObjRef push_object(const Obj* obj);
|
||||
// pull object from a potentially remote object store
|
||||
slice pull_object(ObjRef objref);
|
||||
// stores an object to the local object store
|
||||
void put_obj(ObjRef objref, const Obj* obj);
|
||||
void put_object(ObjRef objref, const Obj* obj);
|
||||
// retrieve serialized object from local object store
|
||||
slice get_object(ObjRef objref);
|
||||
// register function with scheduler
|
||||
void register_function(const std::string& name, size_t num_return_vals);
|
||||
// start the worker server which accepts tasks from the scheduler and stores
|
||||
|
||||
+10
-11
@@ -1,5 +1,5 @@
|
||||
import unittest
|
||||
import orchpy.unison as unison
|
||||
import orchpy
|
||||
import orchpy.services as services
|
||||
import orchpy.worker as worker
|
||||
import numpy as np
|
||||
@@ -82,22 +82,22 @@ class ObjStoreTest(unittest.TestCase):
|
||||
objstore2_stub = connect_to_objstore(IP_ADDRESS, objstore2_port)
|
||||
|
||||
worker1 = worker.Worker()
|
||||
worker1.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, worker1_port), address(IP_ADDRESS, objstore1_port))
|
||||
worker.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, objstore1_port), address(IP_ADDRESS, worker1_port), worker1)
|
||||
|
||||
worker2 = worker.Worker()
|
||||
worker2.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, worker2_port), address(IP_ADDRESS, objstore2_port))
|
||||
worker.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, objstore2_port), address(IP_ADDRESS, worker2_port), worker2)
|
||||
|
||||
# pushing and pulling an object shouldn't change it
|
||||
for data in ["h", "h" * 10000, 0, 0.0]:
|
||||
objref = worker1.push(data)
|
||||
result = worker1.pull(objref)
|
||||
objref = worker.push(data, worker1)
|
||||
result = worker.pull(objref, worker1)
|
||||
self.assertEqual(result, data)
|
||||
|
||||
# pushing an object, shipping it to another worker, and pulling it shouldn't change it
|
||||
for data in ["h", "h" * 10000, 0, 0.0]:
|
||||
objref = worker1.push(data)
|
||||
response = objstore1_stub.DeliverObj(orchestra_pb2.DeliverObjRequest(objref=objref.get_id(), objstore_address=address(IP_ADDRESS, objstore2_port)), TIMEOUT_SECONDS)
|
||||
result = worker2.pull(objref)
|
||||
objref = worker.push(data, worker1)
|
||||
response = objstore1_stub.DeliverObj(orchestra_pb2.DeliverObjRequest(objref=objref.val, objstore_address=address(IP_ADDRESS, objstore2_port)), TIMEOUT_SECONDS)
|
||||
result = worker.pull(objref, worker2)
|
||||
self.assertEqual(result, data)
|
||||
|
||||
services.cleanup()
|
||||
@@ -121,8 +121,7 @@ class SchedulerTest(unittest.TestCase):
|
||||
time.sleep(0.2)
|
||||
|
||||
worker1 = worker.Worker()
|
||||
worker1.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, worker1_port), address(IP_ADDRESS, objstore_port))
|
||||
worker1.start_worker_service()
|
||||
worker.connect(address(IP_ADDRESS, scheduler_port), address(IP_ADDRESS, objstore_port), address(IP_ADDRESS, worker1_port), worker1)
|
||||
|
||||
test_dir = os.path.dirname(os.path.abspath(__file__))
|
||||
test_path = os.path.join(test_dir, "testrecv.py")
|
||||
@@ -130,7 +129,7 @@ class SchedulerTest(unittest.TestCase):
|
||||
|
||||
time.sleep(0.2)
|
||||
|
||||
worker1.call("hello_world", ["hi"])
|
||||
worker1.remote_call("print_string", ["hi"])
|
||||
|
||||
time.sleep(0.1)
|
||||
|
||||
|
||||
+25
-4
@@ -1,4 +1,5 @@
|
||||
import orchpy.unison as unison
|
||||
import argparse
|
||||
|
||||
import orchpy.services as services
|
||||
import orchpy.worker as worker
|
||||
|
||||
@@ -6,13 +7,33 @@ from grpc.beta import implementations
|
||||
import orchestra_pb2
|
||||
import types_pb2
|
||||
|
||||
parser = argparse.ArgumentParser(description='Parse addresses for the worker to connect to.')
|
||||
parser.add_argument("--ip_address", default="127.0.0.1", help="the IP address to use for both the scheduler and objstore")
|
||||
parser.add_argument("--scheduler_port", default=10001, type=int, help="the scheduler's port")
|
||||
parser.add_argument("--objstore_port", default=20001, type=int, help="the objstore's port")
|
||||
parser.add_argument("--worker_port", default=40001, type=int, help="the worker's port")
|
||||
|
||||
@worker.distributed([str], [str])
|
||||
def print_string(string):
|
||||
print "called print_string with", string
|
||||
f = open("asdfasdf.txt", "w")
|
||||
f.write("successfully called print_string with argument {}.".format(string))
|
||||
return string
|
||||
|
||||
@worker.distributed([int, int], [int, int])
|
||||
def handle_int(a, b):
|
||||
return a + 1, b + 1
|
||||
|
||||
def connect_to_scheduler(host, port):
|
||||
channel = implementations.insecure_channel(host, port)
|
||||
return orchestra_pb2.beta_create_Scheduler_stub(channel)
|
||||
|
||||
def address(host, port):
|
||||
return host + ":" + str(port)
|
||||
|
||||
if __name__ == '__main__':
|
||||
scheduler_stub = connect_to_scheduler("127.0.0.1", 22221)
|
||||
worker = worker.Worker()
|
||||
worker.connect("127.0.0.1:22221", "127.0.0.1:10000", "127.0.0.1:22222")
|
||||
args = parser.parse_args()
|
||||
scheduler_stub = connect_to_scheduler(args.ip_address, args.scheduler_port)
|
||||
worker.connect(address(args.ip_address, args.scheduler_port), address(args.ip_address, args.objstore_port), address(args.ip_address, args.worker_port))
|
||||
import IPython
|
||||
IPython.embed()
|
||||
|
||||
+18
-21
@@ -1,37 +1,34 @@
|
||||
import sys
|
||||
import argparse
|
||||
|
||||
import orchpy.unison as unison
|
||||
import orchpy
|
||||
import orchpy.services as services
|
||||
import orchpy.worker as worker
|
||||
|
||||
parser = argparse.ArgumentParser(description='Parse addresses for the worker to connect to.')
|
||||
parser.add_argument("--ip_address", default="127.0.0.1", help="the IP address to use for both the scheduler and objstore")
|
||||
parser.add_argument("--scheduler_port", default=10001, type=int, help="the scheduler's port")
|
||||
parser.add_argument("--objstore_port", default=20001, type=int, help="the objstore's port")
|
||||
parser.add_argument("--worker_port", default=40001, type=int, help="the worker's port")
|
||||
|
||||
@worker.distributed([str], [str])
|
||||
def print_string(string):
|
||||
print "called print_string with", string
|
||||
f = open("asdfasdf.txt", "w")
|
||||
f.write("successfully called print_string with argument {}.".format(string))
|
||||
return string
|
||||
|
||||
@worker.distributed([int, int], [int, int])
|
||||
def handle_int(a, b):
|
||||
return a + 1, b + 1
|
||||
|
||||
def address(host, port):
|
||||
return host + ":" + str(port)
|
||||
|
||||
if __name__ == '__main__':
|
||||
ip_address = sys.argv[1]
|
||||
scheduler_port = sys.argv[2]
|
||||
worker_port = sys.argv[3]
|
||||
objstore_port = sys.argv[4]
|
||||
args = parser.parse_args()
|
||||
worker.connect(address(args.ip_address, args.scheduler_port), address(args.ip_address, args.objstore_port), address(args.ip_address, args.worker_port))
|
||||
|
||||
def address(host, port):
|
||||
return host + ":" + str(port)
|
||||
worker.global_worker.register_function(print_string)
|
||||
worker.global_worker.register_function(handle_int)
|
||||
|
||||
worker = worker.Worker()
|
||||
worker.connect(address(ip_address, scheduler_port), address(ip_address, worker_port), address(ip_address, objstore_port))
|
||||
worker.start_worker_service()
|
||||
|
||||
worker.register_function("print_string", print_string, 0)
|
||||
worker.register_function("handle_int", handle_int, 0)
|
||||
|
||||
name, args, returnref = worker.wait_for_next_task()
|
||||
print "received args ", args
|
||||
if args == ["hi"]:
|
||||
sys.exit(0)
|
||||
else:
|
||||
sys.exit(1)
|
||||
worker.main_loop()
|
||||
|
||||
Reference in New Issue
Block a user