Configure scheduler internal/external ip with public_ingress option (#…

…164)

* configure scheduler internal/external ip with public_ingress option

* Black

* do not create public address if public_ingress is false

* Bump GHA setup-miniconda version

* Bump GHA setup-miniconda version for imports too

Co-authored-by: Jacob Tomlinson <jtomlinson@nvidia.com>

.github/workflows/ci.yml

@@ -16,7 +16,7 @@ jobs:
         uses: actions/checkout@v2
 
       - name: Setup Conda Environment
-        uses: goanpeca/setup-miniconda@v1.7.0
+        uses: conda-incubator/setup-miniconda@v2
         with:
           miniconda-version: "latest"
           python-version: ${{ matrix.python-version }}
@@ -47,7 +47,7 @@ jobs:
         uses: actions/checkout@v2
 
       - name: Setup Conda Environment
-        uses: goanpeca/setup-miniconda@v1.7.0
+        uses: conda-incubator/setup-miniconda@v2
         with:
           miniconda-version: "latest"
           python-version: "3.7"

dask_cloudprovider/cloudprovider.yaml

@@ -78,3 +78,4 @@ cloudprovider:
     gpu_type: "" # type of gpus to use: nvidia-tesla-k80, nvidia-tesla-p100, nvidia-tesla-t4
     docker_image: "daskdev/dask:latest" # docker image to use
     auto_shutdown: true # Shutdown instances automatically if the scheduler or worker services time out.
+    public_ingress: true # configure the scheduler to be externally accessible.  This assumes firefwall rules for 8787 and 8786

dask_cloudprovider/gcp/instances.py

@@ -112,14 +112,6 @@ def create_gcp_config(self):
                 {
                     "kind": "compute#networkInterface",
                     "subnetwork": f"projects/{self.projectid}/regions/{self.general_zone}/subnetworks/default",
-                    "accessConfigs": [
-                        {
-                            "kind": "compute#accessConfig",
-                            "name": "External NAT",
-                            "type": "ONE_TO_ONE_NAT",
-                            "networkTier": "PREMIUM",
-                        }
-                    ],
                     "aliasIpRanges": [],
                 }
             ],
@@ -162,6 +154,16 @@ def create_gcp_config(self):
             "reservationAffinity": {"consumeReservationType": "ANY_RESERVATION"},
         }
 
+        if self.config.get("public_ingress", True):
+            config["networkInterfaces"][0]["accessConfigs"] = [
+                {
+                    "kind": "compute#accessConfig",
+                    "name": "External NAT",
+                    "type": "ONE_TO_ONE_NAT",
+                    "networkTier": "PREMIUM",
+                }
+            ]
+
         if self.ngpus:
             config["guestAccelerators"] = [
                 {
@@ -201,7 +203,10 @@ async def create_vm(self):
             await asyncio.sleep(0.5)
 
         self.internal_ip = self.get_internal_ip()
-        self.external_ip = self.get_external_ip()
+        if self.config.get("public_ingress", True):
+            self.external_ip = self.get_external_ip()
+        else:
+            self.external_ip = None
         self.cluster._log(
             f"{self.name}\n\tInternal IP: {self.internal_ip}\n\tExternal IP: {self.external_ip}"
         )
@@ -272,7 +277,13 @@ async def start_scheduler(self):
         )
         self.cluster._log("Creating scheduler instance")
         self.internal_ip, self.external_ip = await self.create_vm()
-        self.address = f"tcp://{self.external_ip}:8786"
+
+        if self.config.get("public_ingress", True):
+            # scheduler is publicly available
+            self.address = f"tcp://{self.external_ip}:8786"
+        else:
+            # scheduler is only accessible within VPC
+            self.address = f"tcp://{self.internal_ip}:8786"
         await self.wait_for_scheduler()
 
         # need to reserve internal IP for workers
@@ -325,7 +336,11 @@ async def start(self):
     async def start_worker(self):
         self.cluster._log("Creating worker instance")
         self.internal_ip, self.external_ip = await self.create_vm()
-        self.address = self.external_ip
+        if self.config.get("public_ingress", True):
+            # scheduler is publicly available
+            self.address = self.external_ip
+        else:
+            self.address = self.internal_ip
 
 
 class GCPCluster(VMCluster):