Integrate kbatch

qhub/initialize.py

@@ -58,6 +58,9 @@
     "monitoring": {
         "enabled": True,
     },
+    "kbatch": {
+        "enabled": True,
+    },
     "cdsdashboards": {
         "enabled": True,
         "cds_hide_user_named_servers": True,

qhub/schema.py

@@ -74,6 +74,13 @@ class HelmExtension(Base):
     overrides: typing.Optional[typing.Dict]
 
 
+# ============== kbatch =============
+
+
+class KBatch(Base):
+    enabled: bool
+
+
 # ============== Monitoring =============
 
 
@@ -452,6 +459,7 @@ class Main(Base):
     theme: Theme
     profiles: Profiles
     environments: typing.Dict[str, CondaEnvironment]
+    kbatch: typing.Optional[KBatch]
     monitoring: typing.Optional[Monitoring]
     clearml: typing.Optional[ClearML]
     tf_extensions: typing.Optional[typing.List[QHubExtension]]

qhub/stages/input_vars.py

@@ -251,6 +251,7 @@ def stage_07_kubernetes_services(stage_outputs, config):
         "dask-gateway-profiles": config["profiles"]["dask_worker"],
         # monitoring
         "monitoring-enabled": config["monitoring"]["enabled"],
+        "kbatch-enabled": config["kbatch"]["enabled"],
         # prefect
         "prefect-enabled": config.get("prefect", {}).get("enabled", False),
         "prefect-token": config.get("prefect", {}).get("token", ""),

qhub/template/stages/07-kubernetes-services/jupyterhub.tf

@@ -112,7 +112,10 @@ module "jupyterhub" {
 
   services = concat([
     "dask-gateway"
-  ], (var.prefect-enabled ? ["prefect"] : []))
+    ],
+    (var.prefect-enabled ? ["prefect"] : []),
+    (var.kbatch-enabled ? ["kbatch"] : [])
+  )
 
   general-node-group = var.node_groups.general
   user-node-group  = var.node_groups.user

qhub/template/stages/07-kubernetes-services/kbatch.tf

@@ -0,0 +1,23 @@
+# ======================= VARIABLES ======================
+variable "kbatch-enabled" {
+  description = "kbatch enabled or disabled"
+  type = bool
+}
+
+
+# ====================== RESOURCES =======================
+module "kbatch" {
+  count = var.kbatch-enabled ? 1 : 0
+
+  source = "./modules/kubernetes/services/kbatch"
+
+  namespace            = var.environment
+  external-url         = var.endpoint
+
+  jupyterhub_api_token = module.jupyterhub.services.kbatch.api_token
+  node-group   = var.node_groups.user
+
+  dask-gateway-address = module.dask-gateway.config.gateway.address
+  dask-gateway-proxy-address = module.dask-gateway.config.gateway.proxy_address
+  dask-worker-image = var.dask-worker-image
+}

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/.ipynb_checkpoints/kbatch_dask_gateway_job-checkpoint.yaml

@@ -0,0 +1,6 @@
+name: dask-gateway-job
+image: mcr.microsoft.com/planetary-computer/python:2021.08.02.0
+args:
+  - python
+  - kbatch_dask_gateway_test.py
+code: kbatch_dask_gateway_test.py

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/.ipynb_checkpoints/kbatch_nb_job-checkpoint.yaml

@@ -0,0 +1,6 @@
+name: nb-job
+image: mcr.microsoft.com/planetary-computer/python:latest
+args:
+  - papermill
+  - kbatch_nb_test.ipynb
+code: kbatch_nb_test.ipynb

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/kbatch_dask_gateway.py

@@ -0,0 +1,37 @@
+"""
+Start a cluster with Dask Gateway, print the dashboard link, and run some tasks.
+"""
+import dask
+import dask_gateway
+import distributed
+from distributed import wait
+
+
+def inc(x):
+    return x + 1
+
+
+def main():
+    print(f"dask version         = {dask.__version__}")
+    print(f"dask_gateway version = {dask_gateway.__version__}")
+    print(f"distributed version  = {distributed.__version__}")
+
+    gateway = dask_gateway.Gateway()
+    options = gateway.cluster_options(use_local_defaults=False)
+
+    print("Starting cluster")
+    cluster = gateway.new_cluster(options)
+    client = cluster.get_client()
+    print("Dashboard at:", client.dashboard_link)
+
+    cluster.scale(2)
+
+    futures = client.map(inc, list(range(100)))
+    _ = wait(futures)
+
+    print("Closing cluster")
+    cluster.close()
+
+
+if __name__ == "__main__":
+    main()

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/kbatch_dask_gateway_job.yaml

@@ -0,0 +1,6 @@
+name: dask-gateway-job
+image: mcr.microsoft.com/planetary-computer/python:2021.08.02.0
+args:
+  - python
+  - kbatch_dask_gateway.py
+code: kbatch_dask_gateway.py

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/kbatch_nb.ipynb

@@ -0,0 +1,121 @@
+{
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "cba309a0-d8ff-4c18-8d94-c380e875ac46",
+   "metadata": {
+    "tags": []
+   },
+   "source": [
+    "# Canonical Tests\n",
+    "\n",
+    "The goal of the work we performed on behalf of OGC/USGS was to enable users on QHub (and perhaps plain JupyterHub) to:\n",
+    "- [ ] run long-running notebooks or scripts\n",
+    "- [ ] run notebooks and scripts as cronjobs\n",
+    "\n",
+    "And as a stretch goal:\n",
+    "- [ ] run complex workflows that require multiple steps / noteboks or scripts\n",
+    "\n",
+    "Additional requirements:\n",
+    "- the notebook or script should work even after the user's JupyterLab session ends\n",
+    "- the notebook or script can connect to the Dask-Gateway and launch a Dask cluster\n",
+    "\n",
+    "This notebooks will serve as a \"unit test\" for the above features, for more details [see this issue](https://github.com/Quansight/ogc-management/issues/6).\n",
+    "\n",
+    "## `kbatch`\n",
+    "\n",
+    "The first two features outline above will be handled by [`kbatch`](https://github.com/kbatch-dev/kbatch). `kbatch` consists of two major components, the frontend `kbatch` and backend `kbatch-proxy`. The user submits job requests to `kbatch-proxy` and `kbatch-proxy` submits those job requests to the Kubernetes API. \n",
+    "\n",
+    "> NOTE:\n",
+    "> At the present, no additional features have been added to `kbatch`, we have simply integrated `kbatch-proxy` into QHub. A feature enhancement PR will need to be opened on the `kbatch` repo in order to enable cronjobs.\n",
+    "\n",
+    "### Setup\n",
+    "\n",
+    "In order to use `kbatch` in it's current form, some basic setup is required of the user. Going forward, we will assume that `kbatch-proxy` has been correctly integrated into QHub.\n",
+    "\n",
+    "1. Create or modify a conda environment by adding `kbatch`. And activate this conda environment.\n",
+    "\n",
+    "```\n",
+    "pip install kbatch\n",
+    "```\n",
+    "\n",
+    "2. Create a JupyterHub API token and configure `kbatch` to talk to the `kbatch-proxy` service.\n",
+    "\n",
+    "```\n",
+    "kbatch configure --kbatch-url http://kbatch-kbatch-proxy.dev.svc.cluster.local --token <JUPYTERHUB_API_TOKEN>\n",
+    "```\n",
+    "\n",
+    "3. Submit a job to `kbatch`\n",
+    "\n",
+    "```\n",
+    "kbatch job submit --name=list-files --image=alpine --command='[\"ls\", \"-lh\"]'\n",
+    "```\n",
+    "\n",
+    "### Run this notebook\n",
+    "\n",
+    "To run this notebook as a job, you will need an image with `papermill` (or a similar CLI tool). \n",
+    "\n",
+    "Create a configuration file, `kbatch_nb_job.yaml` like the one below:\n",
+    "```yaml\n",
+    "# filename: kbatch_nb_job.yaml\n",
+    "name: nb-job\n",
+    "image: mcr.microsoft.com/planetary-computer/python:latest\n",
+    "args:\n",
+    "  - papermill\n",
+    "  - kbatch_nb.ipynb\n",
+    "code: kbatch_nb.ipynb\n",
+    "```\n",
+    "\n",
+    "Then run:\n",
+    "```\n",
+    "kbatch job submit -f kbatch_nb_job.yaml\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 9,
+   "id": "ff3e90f6-f0b6-4300-81c2-b7ade75d57b4",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import time"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 11,
+   "id": "277ff7d7-d938-4790-af92-5719534b08d5",
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "with open('kbatch_nb_output.txt', 'w') as f:\n",
+    "    for i in range(0,10):\n",
+    "        current_time = time.strftime(\"%Y-%m-%d-%H:%M:%S\", time.localtime())\n",
+    "        time.sleep(1)\n",
+    "        f.write(f'{current_time}: {i}\\n')"
+   ]
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Python 3 (ipykernel)",
+   "language": "python",
+   "name": "python3"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.10.4"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 5
+}

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/jupyterhub/files/examples/kbatch_nb_job.yaml

@@ -0,0 +1,6 @@
+name: nb-job
+image: mcr.microsoft.com/planetary-computer/python:latest
+args:
+  - papermill
+  - kbatch_nb.ipynb
+code: kbatch_nb.ipynb

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/kbatch/main.tf

@@ -0,0 +1,81 @@
+# Get this name dynamtically
+locals {
+  kbatch_service_account_name = "kbatch-kbatch-proxy"
+}
+
+resource "helm_release" "kbatch" {
+  name       = "kbatch"
+  namespace  = var.namespace
+  repository = "https://kbatch-dev.github.io/helm-chart"
+  chart      = "kbatch-proxy"
+  version    = "0.3.1"
+
+  values = concat([
+    file("${path.module}/values.yaml"),
+    jsonencode({
+      app = {
+        jupyterhub_api_token = var.jupyterhub_api_token
+        jupyterhub_api_url = "https://${var.external-url}/hub/api/"
+        extra_env = {
+          KBATCH_PREFIX = ""
+          KBATCH_JOB_EXTRA_ENV = jsonencode({
+            DASK_GATEWAY__AUTH__TYPE = "jupyterhub"
+            DASK_GATEWAY__CLUSTER__OPTIONS__IMAGE = "${var.dask-worker-image.name}:${var.dask-worker-image.tag}"
+            DASK_GATEWAY__ADDRESS = "${var.dask-gateway-address}"
+            DASK_GATEWAY__PROXY_ADDRESS = "${var.dask-gateway-proxy-address}"
+          })
+        }
+      }
+      image = {
+          tag = "0.3.1"
+      }
+    })
+  ])
+
+  set_sensitive {
+    name  = "jupyterHubToken"
+    value = var.jupyterhub_api_token
+  }
+
+  set {
+    name  = "kbatchImage"
+    value = var.image
+  }
+
+  set {
+    name  = "namespace"
+    value = var.namespace
+  }
+
+}
+
+resource "kubernetes_cluster_role" "kbatch" {
+  metadata {
+    name = "${var.name}-kbatch"
+  }
+
+  rule {
+    api_groups = ["", "batch"]
+    resources  = ["*"]
+    verbs      = ["get", "watch", "list", "patch", "create"]
+  }
+}
+
+
+resource "kubernetes_cluster_role_binding" "kbatch" {
+  metadata {
+    name = "${var.name}-kbatch"
+  }
+
+  role_ref {
+    api_group = "rbac.authorization.k8s.io"
+    kind      = "ClusterRole"
+    name      = kubernetes_cluster_role.kbatch.metadata.0.name
+  }
+  subject {
+    kind      = "ServiceAccount"
+    name      = local.kbatch_service_account_name
+    namespace = var.namespace
+    api_group = ""
+  }
+}

qhub/template/stages/07-kubernetes-services/modules/kubernetes/services/kbatch/values.yaml

@@ -0,0 +1 @@
+# https://github.com/kbatch-dev/helm-chart/blob/main/kbatch/values.yaml