Removing custom dask_gateway from qhub and idle_timeout for dask clus…

…ters to 30 min (#1151) * Removing custom dask_gateway from qhub Closes #1012 * Adding idle timeout to dask gateway * Fixing broken links
nebari-dev · Mar 22, 2022 · d4cc037 · d4cc037
1 parent ec0a9ce
commit d4cc037
Show file tree

Hide file tree

Showing 13 changed files with 11 additions and 118 deletions.
diff --git a/.github/workflows/image.yaml b/.github/workflows/image.yaml
@@ -27,7 +27,6 @@ jobs:
           - jupyterlab
           - jupyterhub
           - dask-worker
-          - dask-gateway
     steps:
       - name: 'Checkout Infrastructure'
         uses: actions/checkout@main

diff --git a/docs/source/installation/configuration.md b/docs/source/installation/configuration.md
@@ -450,7 +450,7 @@ local:
 ## Terraform state
 
 Terraform manages the state of all the deployed resources via
-[backends](https://www.terraform.io/docs/language/settings/backends/index.html). Terraform
+[backends](https://www.terraform.io/language/settings/backends). Terraform
 requires storing the state in order to keep track of the names, ids,
 and states of deployed resources. The simplest approach is storing the
 state on the local filesystem but isn't recommended and isn't the
@@ -459,7 +459,7 @@ default of QHub. `terraform_state` is either `remote`, `existing` or
 control the state of the cluster `local` via tfstate file (not
 recommended), on an already `existing` terraform state store or
 remotely and auto creating the terraform state store. See [terraform
-remote state](https://www.terraform.io/docs/language/state/index.html)
+remote state](https://www.terraform.io/language/state/remote)
 docs. If you are doing anything other than testing we highly recommend
 `remote` unless you know what you are doing.
 
@@ -495,15 +495,14 @@ terraform_state:
 Default images are to the default image run if not specified in a
 profile (described in the next section). The `jupyterhub` key controls
 the jupyterhub image run. These control the docker image used to run
-JupyterHub, the default JupyterLab image, the default Dask worker
-image, and Dask Gateway docker image.
+JupyterHub, the default JupyterLab image, and the default Dask worker
+image.
 
 ```yaml
 default_images:
   jupyterhub: "quansight/qhub-jupyterhub:v||QHUB_VERSION||"
   jupyterlab: "quansight/qhub-jupyterlab:v||QHUB_VERSION||"
   dask_worker: "quansight/qhub-dask-worker:v||QHUB_VERSION||"
-  dask_gateway: "quansight/qhub-dask-gateway:v||QHUB_VERSION||"
 ```
 
 ## Storage

diff --git a/docs/source/installation/existing.md b/docs/source/installation/existing.md
@@ -128,8 +128,6 @@ default_images:
   jupyterhub: quansight/qhub-jupyterhub:v0.3.13
   jupyterlab: quansight/qhub-jupyterlab:v0.3.13
   dask_worker: quansight/qhub-dask-worker:v0.3.13
-  dask_gateway: quansight/qhub-dask-gateway:v0.3.13
-  conda_store: quansight/conda-store-server:v0.3.3
 storage:
   conda_store: 60Gi
   shared_filesystem: 100Gi

diff --git a/qhub/initialize.py b/qhub/initialize.py
@@ -37,7 +37,6 @@
         "jupyterhub": f"quansight/qhub-jupyterhub:{qhub_image_tag}",
         "jupyterlab": f"quansight/qhub-jupyterlab:{qhub_image_tag}",
         "dask_worker": f"quansight/qhub-dask-worker:{qhub_image_tag}",
-        "dask_gateway": f"quansight/qhub-dask-gateway:{qhub_image_tag}",
     },
     "storage": {"conda_store": "60Gi", "shared_filesystem": "100Gi"},
     "theme": {

diff --git a/qhub/schema.py b/qhub/schema.py
@@ -121,7 +121,6 @@ class DefaultImages(Base):
     jupyterhub: str
     jupyterlab: str
     dask_worker: str
-    dask_gateway: str
 
 
 # =========== Authentication ==============

diff --git a/qhub/stages/input_vars.py b/qhub/stages/input_vars.py
@@ -244,9 +244,6 @@ def stage_07_kubernetes_services(stage_outputs, config):
             .get("extraEnv", [])
         ),
         # dask-gateway
-        "dask-gateway-image": _split_docker_image_name(
-            config["default_images"]["dask_gateway"]
-        ),
         "dask-worker-image": _split_docker_image_name(
             config["default_images"]["dask_worker"]
         ),

diff --git a/qhub/template/image/Dockerfile.dask-gateway b/qhub/template/image/Dockerfile.dask-gateway
diff --git a/qhub/template/image/docker-bake.json b/qhub/template/image/docker-bake.json
@@ -7,23 +7,13 @@
   "group": {
     "default": {
       "targets": [
-        "dask-gateway",
         "dask-worker",
         "jupyterhub",
         "jupyterlab"
       ]
     }
   },
   "target": {
-    "dask-gateway": {
-      "context": "./",
-      "dockerfile": "Dockerfile.dask-gateway",
-      "tags": [
-        "docker.io/quansight/qhub-dask-gateway:${TAG}",
-        "ghcr.io/quansight/qhub-dask-gateway:${TAG}",
-        "quay.io/quansight/qhub-dask-gateway:${TAG}"
-      ]
-    },
     "dask-worker": {
       "context": "./",
       "dockerfile": "Dockerfile.dask-worker",

diff --git a/qhub/template/stages/07-kubernetes-services/dask_gateway.tf b/qhub/template/stages/07-kubernetes-services/dask_gateway.tf
@@ -7,14 +7,6 @@ variable "dask-worker-image" {
   })
 }
 
-variable "dask-gateway-image" {
-  description = "Dask worker image"
-  type = object({
-    name = string
-    tag  = string
-  })
-}
-
 variable "dask-gateway-profiles" {
   description = "Dask Gateway profiles to expose to user"
   default = []
@@ -32,8 +24,6 @@ module "dask-gateway" {
   external-url = var.endpoint
 
   cluster-image    = var.dask-worker-image
-  gateway-image    = var.dask-gateway-image
-  controller-image = var.dask-gateway-image
 
   general-node-group = var.node_groups.general
   worker-node-group  = var.node_groups.worker

diff --git a/...7-kubernetes-services/modules/kubernetes/services/dask-gateway/files/controller_config.py b/...7-kubernetes-services/modules/kubernetes/services/dask-gateway/files/controller_config.py
@@ -32,38 +32,3 @@ def dask_gateway_config(path="/var/lib/dask-gateway/config.json"):
 
 c.KubeController.proxy_web_entrypoint = "websecure"
 c.KubeController.proxy_tcp_entrypoint = "tcp"
-
-
-# override default make_ingressroute to add tls cert resolver
-def make_ingressroute(self, cluster_name, namespace):
-    route = f"{self.proxy_prefix}/clusters/{namespace}.{cluster_name}/"
-    return {
-        "apiVersion": "traefik.containo.us/v1alpha1",
-        "kind": "IngressRoute",
-        "metadata": {
-            "labels": self.get_labels(cluster_name, "dask-scheduler"),
-            "annotations": self.common_annotations,
-            "name": f"dask-{cluster_name}",
-        },
-        "spec": {
-            "entryPoints": [self.proxy_web_entrypoint],
-            "routes": [
-                {
-                    "kind": "Rule",
-                    "match": f"PathPrefix(`{route}`)",
-                    "services": [
-                        {
-                            "name": self.make_service_name(cluster_name),
-                            "namespace": namespace,
-                            "port": 8787,
-                        }
-                    ],
-                    "middlewares": self.proxy_web_middlewares,
-                }
-            ],
-            "tls": {},
-        },
-    }
-
-
-c.KubeController.make_ingressroute = make_ingressroute
diff --git a/...s/07-kubernetes-services/modules/kubernetes/services/dask-gateway/files/gateway_config.py b/...s/07-kubernetes-services/modules/kubernetes/services/dask-gateway/files/gateway_config.py
@@ -30,6 +30,7 @@ def dask_gateway_config(path="/var/lib/dask-gateway/config.json"):
 )
 c.KubeClusterConfig.image_pull_policy = config["cluster"]["image_pull_policy"]
 c.KubeClusterConfig.environment = config["cluster"]["environment"]
+c.KubeClusterConfig.idle_timeout = config["cluster"]["idle_timeout"]
 
 c.KubeClusterConfig.scheduler_cores = config["cluster"]["scheduler_cores"]
 c.KubeClusterConfig.scheduler_cores_limit = config["cluster"]["scheduler_cores_limit"]

diff --git a/...plate/stages/07-kubernetes-services/modules/kubernetes/services/dask-gateway/variables.tf b/...plate/stages/07-kubernetes-services/modules/kubernetes/services/dask-gateway/variables.tf
@@ -144,6 +144,7 @@ variable "cluster" {
     worker_extra_container_config = any
     worker_extra_pod_config       = any
     # additional fields
+    idle_timeout      = number
     image_pull_policy = string
     environment       = map(string)
   })
@@ -163,6 +164,7 @@ variable "cluster" {
     worker_extra_container_config = {}
     worker_extra_pod_config       = {}
     # additional fields
+    idle_timeout      = 1800 # 30 minutes
     image_pull_policy = "IfNotPresent"
     environment       = {}
   }

diff --git a/qhub/upgrade.py b/qhub/upgrade.py
@@ -318,6 +318,10 @@ def _version_specific_upgrade(
         if "conda_store" in config["default_images"]:
             del config["default_images"]["conda_store"]
 
+        # Remove dask_gateway image from default_images
+        if "dask_gateway" in config["default_images"]:
+            del config["default_images"]["dask_gateway"]
+
         # Create root password
         default_password = "".join(
             secrets.choice(string.ascii_letters + string.digits) for i in range(16)