Make GraphScope support ResourceQuota limit in namespace (#102)

coordinator/gscoordinator/cluster.py

@@ -103,6 +103,12 @@ def __init__(
         gie_graph_manager_image=None,
         coordinator_name=None,
         coordinator_service_name=None,
+        etcd_cpu=None,
+        etcd_mem=None,
+        zookeeper_cpu=None,
+        zookeeper_mem=None,
+        gie_graph_manager_cpu=None,
+        gie_graph_manager_mem=None,
         engine_cpu=None,
         engine_mem=None,
         vineyard_cpu=None,
@@ -165,12 +171,18 @@ def __init__(
 
         # etcd pod info
         self._etcd_image = etcd_image
+        self._etcd_cpu = etcd_cpu
+        self._etcd_mem = etcd_mem
 
         # zookeeper pod info
         self._zookeeper_image = zookeeper_image
+        self._zookeeper_cpu = zookeeper_cpu
+        self._zookeeper_mem = zookeeper_mem
 
         # interactive engine graph manager info
         self._gie_graph_manager_image = gie_graph_manager_image
+        self._gie_graph_manager_cpu = gie_graph_manager_cpu
+        self._gie_graph_manager_mem = gie_graph_manager_mem
 
         self._image_pull_policy = image_pull_policy
 
@@ -325,6 +337,8 @@ def _create_etcd(self):
             name=self._etcd_container_name,
             service_name=self._etcd_service_name,
             image=self._etcd_image,
+            cpu=self._etcd_cpu,
+            mem=self._etcd_mem,
             listen_peer_service_port=self._random_etcd_listen_peer_service_port,
             listen_client_service_port=self._random_etcd_listen_client_service_port,
         )
@@ -446,13 +460,17 @@ def _create_interactive_engine_service(self):
         graph_manager_builder.add_manager_container(
             name=self._gie_manager_container_name,
             image=self._gie_graph_manager_image,
+            cpu=self._gie_graph_manager_cpu,
+            mem=self._gie_graph_manager_mem,
             port=self._interactive_engine_manager_port,
         )
 
         # add zookeeper container
         graph_manager_builder.add_zookeeper_container(
             name=self._gie_zookeeper_container_name,
             image=self._zookeeper_image,
+            cpu=self._zookeeper_cpu,
+            mem=self._zookeeper_mem,
             port=self._zookeeper_port,
         )
 

coordinator/gscoordinator/coordinator.py

@@ -416,6 +416,8 @@ def CreateInteractiveInstance(self, request, context):
             request.object_id,
         )
         object_id = request.object_id
+        gremlin_server_cpu = request.gremlin_server_cpu
+        gremlin_server_mem = request.gremlin_server_mem
         manager_host = MAXGRAPH_MANAGER_HOST % (
             self._gie_graph_manager_service_name,
             self._k8s_namespace,
@@ -428,6 +430,8 @@ def CreateInteractiveInstance(self, request, context):
                 "schemaJson": schema_json,
                 "podNameList": ",".join(self._pods_list),
                 "containerName": ENGINE_CONTAINER,
+                "gremlinServerCpu": str(gremlin_server_cpu),
+                "gremlinServerMem": gremlin_server_mem,
             }
             post_data = urllib.parse.urlencode(params).encode("utf-8")
             create_res = urllib.request.urlopen(url=post_url, data=post_data)
@@ -774,6 +778,42 @@ def parse_sys_args():
         default="256Mi",
         help="Memory of engine container, suffix with ['Mi', 'Gi', 'Ti'].",
     )
+    parser.add_argument(
+        "--k8s_etcd_cpu",
+        type=float,
+        default=1.0,
+        help="Cpu cores of etcd pod, default: 1.0",
+    )
+    parser.add_argument(
+        "--k8s_etcd_mem",
+        type=str,
+        default="256Mi",
+        help="Memory of etcd pod, suffix with ['Mi', 'Gi', 'Ti'].",
+    )
+    parser.add_argument(
+        "--k8s_zookeeper_cpu",
+        type=float,
+        default=1.0,
+        help="Cpu cores of zookeeper container, default: 1.0",
+    )
+    parser.add_argument(
+        "--k8s_zookeeper_mem",
+        type=str,
+        default="256Mi",
+        help="Memory of zookeeper container, suffix with ['Mi', 'Gi', 'Ti'].",
+    )
+    parser.add_argument(
+        "--k8s_gie_graph_manager_cpu",
+        type=float,
+        default=1.0,
+        help="Cpu cores of graph manager container, default: 1.0",
+    )
+    parser.add_argument(
+        "--k8s_gie_graph_manager_mem",
+        type=str,
+        default="256Mi",
+        help="Memory of graph manager container, suffix with ['Mi', 'Gi', 'Ti'].",
+    )
     parser.add_argument(
         "--k8s_volumes",
         type=str,
@@ -819,6 +859,12 @@ def launch_graphscope():
             gie_graph_manager_image=args.k8s_gie_graph_manager_image,
             coordinator_name=args.k8s_coordinator_name,
             coordinator_service_name=args.k8s_coordinator_service_name,
+            etcd_cpu=args.k8s_etcd_cpu,
+            etcd_mem=args.k8s_etcd_mem,
+            zookeeper_cpu=args.k8s_zookeeper_cpu,
+            zookeeper_mem=args.k8s_zookeeper_mem,
+            gie_graph_manager_cpu=args.k8s_gie_graph_manager_cpu,
+            gie_graph_manager_mem=args.k8s_gie_graph_manager_mem,
             engine_cpu=args.k8s_engine_cpu,
             engine_mem=args.k8s_engine_mem,
             vineyard_cpu=args.k8s_vineyard_cpu,

interactive_engine/deploy/docker/dockerfile/create_maxgraph_instance.sh

@@ -37,6 +37,7 @@ function _create_pod {
     kubectl create configmap $config_name --from-file /home/maxgraph/config_$object_id
     sed -e "s/unique_pod_name/$pod_name/g" -e "s/unique_config_name/$config_name/g" \
         -e "s/gremlin_image/$gremlin_image/g" -e "s/unique_object_id/$object_id/g" \
+        -e "s/gremlin_server_cpu/$gremlin_server_cpu/g" -e "s/gremlin_server_mem/$gremlin_server_mem/g" \
         -e "s/coordinator_image/$coordinator_image/g" \
         /root/maxgraph/pod.yaml > /root/maxgraph/pod_${object_id}.yaml
     kubectl apply -f /root/maxgraph/pod_${object_id}.yaml
@@ -98,6 +99,8 @@ export schema_path=$2
 export engine_count=`echo $3 | awk -F"," '{print NF}'`
 export pod_hosts=`echo $3 | awk -F"," '{for(i=1;i<=NF;++i) {print $i" "}}'`
 export ENGINE_CONTAINER=$4
-export engine_paras=$5 
+export gremlin_server_cpu=$5
+export gremlin_server_mem=$6
+export engine_paras=$7
 export launch_engine_cmd="export object_id=${object_id} && /home/maxgraph/executor-entrypoint.sh"
 _create_maxgraph_instance

interactive_engine/deploy/docker/dockerfile/pod.yaml

@@ -16,6 +16,13 @@ spec:
   #           - k8s_node_host 
   containers:
   - image: gremlin_image
+    resources:
+      requests:
+        cpu: gremlin_server_cpu
+        memory: gremlin_server_mem
+      limits:
+        cpu: gremlin_server_cpu
+        memory: gremlin_server_mem
     name: gremlin-container
     imagePullPolicy: IfNotPresent
     args: ["/home/maxgraph/frontend-entrypoint.sh"]
@@ -26,6 +33,13 @@ spec:
     - mountPath: /home/maxgraph/config
       name: config-volume
   - image: coordinator_image
+    resources:
+      requests:
+        cpu: gremlin_server_cpu
+        memory: gremlin_server_mem
+      limits:
+        cpu: gremlin_server_cpu
+        memory: gremlin_server_mem
     name: coordinator-container
     imagePullPolicy: IfNotPresent
     args: ["/home/maxgraph/coordinator-entrypoint.sh"]

interactive_engine/src/instance-manager/src/main/java/com/alibaba/maxgraph/admin/controller/InstanceManagerController.java

@@ -92,7 +92,9 @@ private void checkContainerName(String containerName) {
     public CreateInstanceEntity createInstance(@RequestParam("graphName") String graphName,
                                                @RequestParam("schemaJson") String schemaJson,
                                                @RequestParam("podNameList") String podNameList,
-                                               @RequestParam("containerName") String containerName) throws Exception {
+                                               @RequestParam("containerName") String containerName,
+                                               @RequestParam("gremlinServerCpu") String gremlinServerCpu,
+                                               @RequestParam("gremlinServerMem") String gremlinServerMem) throws Exception {
         CreateInstanceEntity createInstanceEntity = new CreateInstanceEntity();
         int errorCode;
         String errorMessage;
@@ -123,6 +125,8 @@ public CreateInstanceEntity createInstance(@RequestParam("graphName") String gra
             createCommandList.add(schemaPath);
             createCommandList.add(podNameList);
             createCommandList.add(containerName);
+            createCommandList.add(gremlinServerCpu);
+            createCommandList.add(gremlinServerMem);
             String command = StringUtils.join(createCommandList, " ");
             logger.info("start to create instance with command " + command);
             Process process = Runtime.getRuntime().exec(command);

proto/message.proto

@@ -141,6 +141,9 @@ message CreateInteractiveRequest {
   // object id of the graph in vineyard
   int64 object_id    = 1;
   string schema_path = 2;
+  // resource
+  float gremlin_server_cpu = 3;
+  string gremlin_server_mem = 4;
 }
 
 message CreateInteractiveResponse {

python/graphscope/client/rpc.py

@@ -148,9 +148,14 @@ def send_heartbeat(self):
         return self._stub.HeartBeat(request)
 
     @catch_grpc_error
-    def create_interactive_engine(self, object_id, schema_path):
+    def create_interactive_engine(
+        self, object_id, schema_path, gremlin_server_cpu, gremlin_server_mem
+    ):
         request = message_pb2.CreateInteractiveRequest(
-            object_id=object_id, schema_path=schema_path
+            object_id=object_id,
+            schema_path=schema_path,
+            gremlin_server_cpu=gremlin_server_cpu,
+            gremlin_server_mem=gremlin_server_mem,
         )
         response = self._stub.CreateInteractiveInstance(request)
         return check_grpc_response(response)

python/graphscope/client/session.py

@@ -150,6 +150,12 @@ def __init__(
         k8s_image_pull_secrets=gs_config.k8s_image_pull_secrets,
         k8s_coordinator_cpu=gs_config.k8s_coordinator_cpu,
         k8s_coordinator_mem=gs_config.k8s_coordinator_mem,
+        k8s_etcd_cpu=gs_config.k8s_etcd_cpu,
+        k8s_etcd_mem=gs_config.k8s_etcd_mem,
+        k8s_zookeeper_cpu=gs_config.k8s_zookeeper_cpu,
+        k8s_zookeeper_mem=gs_config.k8s_zookeeper_mem,
+        k8s_gie_graph_manager_cpu=gs_config.k8s_gie_graph_manager_cpu,
+        k8s_gie_graph_manager_mem=gs_config.k8s_gie_graph_manager_mem,
         k8s_vineyard_cpu=gs_config.k8s_vineyard_cpu,
         k8s_vineyard_mem=gs_config.k8s_vineyard_mem,
         k8s_vineyard_shared_mem=gs_config.k8s_vineyard_shared_mem,
@@ -205,6 +211,22 @@ def __init__(
 
             k8s_coordinator_mem (str, optional): Minimum number of memory request for coordinator pod. Defaults to '4Gi'.
 
+            k8s_etcd_cpu (float, optional): Minimum number of CPU cores request for etcd pod. Defaults to 0.5.
+
+            k8s_etcd_mem (str, optional): Minimum number of memory request for etcd pod. Defaults to '128Mi'.
+
+            k8s_zookeeper_cpu (float, optional):
+                Minimum number of CPU cores request for zookeeper container. Defaults to 0.5.
+
+            k8s_zookeeper_mem (str, optional):
+                Minimum number of memory request for zookeeper container. Defaults to '128Mi'.
+
+            k8s_gie_graph_manager_cpu (float, optional):
+                Minimum number of CPU cores request for graphmanager container. Defaults to 1.0.
+
+            k8s_gie_graph_manager_mem (str, optional):
+                Minimum number of memory request for graphmanager container. Defaults to '4Gi'.
+
             k8s_volumes (dict, optional): A dict of k8s volume which represents a directory containing data, accessible to the
                 containers in a pod. Defaults to {}. Only 'hostPath' supported yet. For example, we can mount host path with:
 
@@ -283,6 +305,12 @@ def __init__(
             "k8s_zookeeper_image",
             "k8s_coordinator_cpu",
             "k8s_coordinator_mem",
+            "k8s_etcd_cpu",
+            "k8s_etcd_mem",
+            "k8s_zookeeper_cpu",
+            "k8s_zookeeper_mem",
+            "k8s_gie_graph_manager_cpu",
+            "k8s_gie_graph_manager_mem",
             "k8s_vineyard_cpu",
             "k8s_vineyard_mem",
             "k8s_vineyard_shared_mem",
@@ -643,6 +671,12 @@ def _connect(self):
                 vineyard_cpu=self._config_params["k8s_vineyard_cpu"],
                 vineyard_mem=self._config_params["k8s_vineyard_mem"],
                 vineyard_shared_mem=self._config_params["k8s_vineyard_shared_mem"],
+                etcd_cpu=self._config_params["k8s_etcd_cpu"],
+                etcd_mem=self._config_params["k8s_etcd_mem"],
+                zookeeper_cpu=self._config_params["k8s_zookeeper_cpu"],
+                zookeeper_mem=self._config_params["k8s_zookeeper_mem"],
+                gie_graph_manager_cpu=self._config_params["k8s_gie_graph_manager_cpu"],
+                gie_graph_manager_mem=self._config_params["k8s_gie_graph_manager_mem"],
                 engine_cpu=self._config_params["k8s_engine_cpu"],
                 engine_mem=self._config_params["k8s_engine_mem"],
                 coordinator_cpu=float(self._config_params["k8s_coordinator_cpu"]),
@@ -859,7 +893,10 @@ def gremlin(self, graph):
         from graphscope.interactive.query import InteractiveQuery
 
         response = self._grpc_client.create_interactive_engine(
-            graph.vineyard_id, graph.schema_path
+            object_id=graph.vineyard_id,
+            schema_path=graph.schema_path,
+            gremlin_server_cpu=gs_config.k8s_gie_gremlin_server_cpu,
+            gremlin_server_mem=gs_config.k8s_gie_gremlin_server_mem,
         )
         interactive_query = InteractiveQuery(
             graphscope_session=self,

python/graphscope/config.py

@@ -43,6 +43,22 @@ class GSConfig(object):
     k8s_coordinator_cpu = 1.0
     k8s_coordinator_mem = "4Gi"
 
+    # etcd resource configuration
+    k8s_etcd_cpu = 0.5
+    k8s_etcd_mem = "128Mi"
+
+    # zookeeper resource configuration
+    k8s_zookeeper_cpu = 0.5
+    k8s_zookeeper_mem = "128Mi"
+
+    # GIE graph manager resource configuration
+    k8s_gie_graph_manager_cpu = 1.0
+    k8s_gie_graph_manager_mem = "4Gi"
+
+    # GIE gremlin server resource configuration
+    k8s_gie_gremlin_server_cpu = 1.0
+    k8s_gie_gremlin_server_mem = "1Gi"
+
     # vineyard resource configuration
     k8s_vineyard_cpu = 0.5
     k8s_vineyard_mem = "512Mi"

python/graphscope/deploy/kubernetes/cluster.py

@@ -90,6 +90,24 @@ class KubernetesCluster(object):
         image_pull_secrets: list of str, optional
             A list of secret name used to pulling image. Defaults to None.
 
+        etcd_cpu: float
+            Minimum number of CPU cores request for etcd pod.
+
+        etcd_mem: str
+             Minimum number of memory request for etcd pod.
+
+        zookeeper_cpu: float
+            Minimum number of CPU cores request for zookeeper container.
+
+        zookeeper_mem: str:
+            Minimum number of memory request for zookeeper container.
+
+        gie_graph_manager_cpu: float
+            Minimum number of CPU cores request for graph manager container.
+
+        gie_graph_manager_mem: str
+            Minimum number of memory request for graph manager container.
+
         vineyard_cpu: float
             Minimum number of CPU cores request for vineyard container.
 
@@ -158,6 +176,12 @@ def __init__(
         engine_mem=None,
         coordinator_cpu=None,
         coordinator_mem=None,
+        etcd_cpu=None,
+        etcd_mem=None,
+        zookeeper_cpu=None,
+        zookeeper_mem=None,
+        gie_graph_manager_cpu=None,
+        gie_graph_manager_mem=None,
         volumes=None,
         timeout_seconds=None,
         waiting_for_delete=None,
@@ -184,6 +208,13 @@ def __init__(
         elif not isinstance(self._image_pull_secrets, list):
             self._image_pull_secrets = [self._image_pull_secrets]
 
+        self._etcd_cpu = etcd_cpu
+        self._etcd_mem = etcd_mem
+        self._zookeeper_cpu = zookeeper_cpu
+        self._zookeeper_mem = zookeeper_mem
+        self._gie_graph_manager_cpu = gie_graph_manager_cpu
+        self._gie_graph_manager_mem = gie_graph_manager_mem
+
         self._vineyard_cpu = vineyard_cpu
         self._vineyard_mem = vineyard_mem
         self._vineyard_shared_mem = vineyard_shared_mem
@@ -422,6 +453,12 @@ def _create_coordinator(self):
             coordinator_cpu=self._coordinator_cpu,
             coordinator_mem=self._coordinator_mem,
             coordinator_service_name=self._coordinator_service_name,
+            etcd_cpu=self._etcd_cpu,
+            etcd_mem=self._etcd_mem,
+            zookeeper_cpu=self._zookeeper_cpu,
+            zookeeper_mem=self._zookeeper_mem,
+            gie_graph_manager_cpu=self._gie_graph_manager_cpu,
+            gie_graph_manager_mem=self._gie_graph_manager_mem,
             vineyard_cpu=self._vineyard_cpu,
             vineyard_mem=self._vineyard_mem,
             vineyard_shared_mem=self._vineyard_shared_mem,