Revert "Updating zero capacity resource semantics (#4555)"

This reverts commit 0f42f87.

Author: devin-petersohn

java/api/src/main/java/org/ray/api/options/BaseTaskOptions.java

@@ -7,19 +7,13 @@
  * The options class for RayCall or ActorCreation.
  */
 public abstract class BaseTaskOptions {
-  public final Map<String, Double> resources;
+  public Map<String, Double> resources;
 
   public BaseTaskOptions() {
     resources = new HashMap<>();
   }
 
   public BaseTaskOptions(Map<String, Double> resources) {
-    for (Map.Entry<String, Double> entry : resources.entrySet()) {
-      if (entry.getValue().compareTo(0.0) <= 0) {
-        throw new IllegalArgumentException(String.format("Resource capacity should be " +
-            "positive, but got resource %s = %f.", entry.getKey(), entry.getValue()));
-      }
-    }
     this.resources = resources;
   }
 

java/example.conf

@@ -14,7 +14,7 @@ ray {
   run-mode = CLUSTER
 
   // Available resources on this node.
-  resources: "CPU:4"
+  resources: "CPU:4,GPU:0"
 
   // The address of the redis server to connect, in format `ip:port`.
   // If not provided, Ray processes will be started locally, including

java/runtime/src/main/java/org/ray/runtime/AbstractRayRuntime.java

@@ -32,6 +32,7 @@
 import org.ray.runtime.task.ArgumentsBuilder;
 import org.ray.runtime.task.TaskLanguage;
 import org.ray.runtime.task.TaskSpec;
+import org.ray.runtime.util.ResourceUtil;
 import org.ray.runtime.util.UniqueIdUtil;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -356,6 +357,11 @@ private TaskSpec createTaskSpec(RayFunc func, PyFunctionDescriptor pyFunctionDes
       resources = new HashMap<>(taskOptions.resources);
     }
 
+    if (!resources.containsKey(ResourceUtil.CPU_LITERAL)
+        && !resources.containsKey(ResourceUtil.CPU_LITERAL.toLowerCase())) {
+      resources.put(ResourceUtil.CPU_LITERAL, 0.0);
+    }
+
     int maxActorReconstruction = 0;
     if (taskOptions instanceof ActorCreationOptions) {
       maxActorReconstruction = ((ActorCreationOptions) taskOptions).maxReconstructions;

java/runtime/src/main/java/org/ray/runtime/config/RayConfig.java

@@ -104,6 +104,10 @@ public RayConfig(Config config) {
             + "setting it to the number of CPU cores: {}", numCpu);
         resources.put("CPU", numCpu * 1.0);
       }
+      if (!resources.containsKey("GPU")) {
+        LOGGER.warn("No GPU resource is set in configuration, setting it to 0");
+        resources.put("GPU", 0.0);
+      }
     }
     // Driver id.
     String driverId = config.getString("ray.driver.id");

java/streaming/src/main/resources/ray.conf

@@ -1,5 +1,5 @@
 ray {
   run-mode = SINGLE_PROCESS
-  resources = "CPU:4"
+  resources = "CPU:4,GPU:0"
   redis.address = ""
 }

java/test/src/main/java/org/ray/api/test/ResourcesManagementTest.java

@@ -46,38 +46,30 @@ public Integer echo(Integer number) {
   @Test
   public void testMethods() {
     TestUtils.skipTestUnderSingleProcess();
-    CallOptions callOptions1 = new CallOptions(ImmutableMap.of("CPU", 4.0));
+    CallOptions callOptions1 = new CallOptions(ImmutableMap.of("CPU", 4.0, "GPU", 0.0));
 
     // This is a case that can satisfy required resources.
     // The static resources for test are "CPU:4,RES-A:4".
     RayObject<Integer> result1 = Ray.call(ResourcesManagementTest::echo, 100, callOptions1);
     Assert.assertEquals(100, (int) result1.get());
 
-    CallOptions callOptions2 = new CallOptions(ImmutableMap.of("CPU", 4.0));
+    CallOptions callOptions2 = new CallOptions(ImmutableMap.of("CPU", 4.0, "GPU", 2.0));
 
     // This is a case that can't satisfy required resources.
     // The static resources for test are "CPU:4,RES-A:4".
     final RayObject<Integer> result2 = Ray.call(ResourcesManagementTest::echo, 200, callOptions2);
     WaitResult<Integer> waitResult = Ray.wait(ImmutableList.of(result2), 1, 1000);
 
-    Assert.assertEquals(1, waitResult.getReady().size());
-    Assert.assertEquals(0, waitResult.getUnready().size());
-
-    try {
-      CallOptions callOptions3 = new CallOptions(ImmutableMap.of("CPU", 0.0));
-      Assert.fail();
-    } catch (RuntimeException e) {
-      // We should receive a RuntimeException indicates that we should not
-      // pass a zero capacity resource.
-    }
+    Assert.assertEquals(0, waitResult.getReady().size());
+    Assert.assertEquals(1, waitResult.getUnready().size());
   }
 
   @Test
   public void testActors() {
     TestUtils.skipTestUnderSingleProcess();
 
     ActorCreationOptions actorCreationOptions1 =
-        new ActorCreationOptions(ImmutableMap.of("CPU", 2.0));
+        new ActorCreationOptions(ImmutableMap.of("CPU", 2.0, "GPU", 0.0));
 
     // This is a case that can satisfy required resources.
     // The static resources for test are "CPU:4,RES-A:4".
@@ -88,7 +80,7 @@ public void testActors() {
     // This is a case that can't satisfy required resources.
     // The static resources for test are "CPU:4,RES-A:4".
     ActorCreationOptions actorCreationOptions2 =
-        new ActorCreationOptions(ImmutableMap.of("CPU", 8.0));
+        new ActorCreationOptions(ImmutableMap.of("CPU", 8.0, "GPU", 0.0));
 
     RayActor<ResourcesManagementTest.Echo> echo2 =
         Ray.createActor(Echo::new, actorCreationOptions2);

python/ray/includes/task.pxi

@@ -44,6 +44,8 @@ cdef class Task:
         # Parse the resource map.
         if resource_map is not None:
             required_resources = resource_map_from_dict(resource_map)
+        if required_resources.count(b"CPU") == 0:
+            required_resources[b"CPU"] = 1.0
         if placement_resource_map is not None:
             required_placement_resources = (
                 resource_map_from_dict(placement_resource_map))

python/ray/services.py

@@ -1029,25 +1029,14 @@ def check_and_update_resources(num_cpus, num_gpus, resources):
         if gpu_ids is not None:
             resources["GPU"] = min(resources["GPU"], len(gpu_ids))
 
-    resources = {
-        resource_label: resource_quantity
-        for resource_label, resource_quantity in resources.items()
-        if resource_quantity != 0
-    }
-
     # Check types.
     for _, resource_quantity in resources.items():
         assert (isinstance(resource_quantity, int)
                 or isinstance(resource_quantity, float))
         if (isinstance(resource_quantity, float)
                 and not resource_quantity.is_integer()):
-            raise ValueError(
-                "Resource quantities must all be whole numbers. Received {}.".
-                format(resources))
-        if resource_quantity < 0:
-            raise ValueError(
-                "Resource quantities must be nonnegative. Received {}.".format(
-                    resources))
+            raise ValueError("Resource quantities must all be whole numbers.")
+
         if resource_quantity > ray_constants.MAX_RESOURCE_QUANTITY:
             raise ValueError("Resource quantities must be at most {}.".format(
                 ray_constants.MAX_RESOURCE_QUANTITY))
@@ -1124,9 +1113,8 @@ def start_raylet(redis_address,
 
     # Limit the number of workers that can be started in parallel by the
     # raylet. However, make sure it is at least 1.
-    num_cpus_static = static_resources.get("CPU", 0)
     maximum_startup_concurrency = max(
-        1, min(multiprocessing.cpu_count(), num_cpus_static))
+        1, min(multiprocessing.cpu_count(), static_resources["CPU"]))
 
     # Format the resource argument in a form like 'CPU,1.0,GPU,0,Custom,3'.
     resource_argument = ",".join(

python/ray/tests/test_basic.py

@@ -1948,15 +1948,15 @@ def run_lots_of_tasks():
     store_names = []
     store_names += [
         client["ObjectStoreSocketName"] for client in client_table
-        if client["Resources"].get("GPU", 0) == 0
+        if client["Resources"]["GPU"] == 0
     ]
     store_names += [
         client["ObjectStoreSocketName"] for client in client_table
-        if client["Resources"].get("GPU", 0) == 5
+        if client["Resources"]["GPU"] == 5
     ]
     store_names += [
         client["ObjectStoreSocketName"] for client in client_table
-        if client["Resources"].get("GPU", 0) == 1
+        if client["Resources"]["GPU"] == 1
     ]
     assert len(store_names) == 3
 
@@ -2126,32 +2126,6 @@ def f():
     ray.get(results)
 
 
-def test_zero_capacity_deletion_semantics(shutdown_only):
-    ray.init(num_cpus=2, num_gpus=1, resources={"test_resource": 1})
-
-    def test():
-        resources = ray.global_state.available_resources()
-        retry_count = 0
-
-        while resources and retry_count < 5:
-            time.sleep(0.1)
-            resources = ray.global_state.available_resources()
-            retry_count += 1
-
-        if retry_count >= 5:
-            raise RuntimeError("Resources were available even after retries.")
-
-        return resources
-
-    function = ray.remote(
-        num_cpus=2, num_gpus=1, resources={"test_resource": 1})(test)
-    cluster_resources = ray.get(function.remote())
-
-    # All cluster resources should be utilized and
-    # cluster_resources must be empty
-    assert cluster_resources == {}
-
-
 @pytest.fixture
 def save_gpu_ids_shutdown_only():
     # Record the curent value of this environment variable so that we can

python/ray/tests/test_global_state.py

@@ -51,8 +51,8 @@ def cpu_task():
 
     while not resource_used:
         available_resources = ray.global_state.available_resources()
-        resource_used = available_resources.get(
-            "CPU", 0) == cluster_resources.get("CPU", 0) - 1
+        resource_used = available_resources[
+            "CPU"] == cluster_resources["CPU"] - 1
 
     assert resource_used