Merge pull request #8 from converged-computing/add-global-token

add support for global token and kind example

.dockerignore

@@ -0,0 +1,2 @@
+rainbow.db
+env

cmd/server/server.go

@@ -10,34 +10,41 @@ import (
 )
 
 var (
-	address     string
+	host        string
 	name        = "rainbow"
 	sqliteFile  = "rainbow.db"
 	environment = "development"
 	cleanup     = false
 	secret      = "chocolate-cookies"
+	globalToken = ""
 )
 
 func main() {
-	flag.StringVar(&address, "address", ":50051", "Server address (host:port)")
+	flag.StringVar(&host, "host", ":50051", "Server address (host:port)")
 	flag.StringVar(&name, "name", name, "Server name (default: rainbow)")
 	flag.StringVar(&sqliteFile, "db", sqliteFile, "sqlite3 database file (default: rainbow.db)")
+	flag.StringVar(&globalToken, "global-token", name, "global token for cluster access (not recommended)")
 	flag.StringVar(&secret, "secret", secret, "secret to validate registration (default: chocolate-cookies)")
 	flag.StringVar(&environment, "environment", environment, "environment (default: development)")
 	flag.BoolVar(&cleanup, "cleanup", cleanup, "cleanup previous sqlite database (default: false)")
 	flag.Parse()
 
 	// create server
 	log.Print("creating 🌈️ server...")
-	s, err := server.NewServer(name, types.Version, environment, sqliteFile, cleanup, secret)
+	s, err := server.NewServer(name, types.Version, environment, sqliteFile, cleanup, secret, globalToken)
 	if err != nil {
 		log.Fatalf("error while creating server: %v", err)
 	}
 	defer s.Stop()
 
+	// Give a warning if the globalToken is set
+	if globalToken != "" {
+		log.Printf("⚠️ WARNING: global-token is set, use with caution.")
+	}
+
 	// run server
 	log.Printf("starting scheduler server: %s", s.String())
-	if err := s.Start(context.Background(), address); err != nil {
+	if err := s.Start(context.Background(), host); err != nil {
 		log.Fatalf("error while running scheduler server: %v", err)
 	}
 	log.Printf("🌈️ done 🌈️")

docs/examples/README.md

@@ -3,4 +3,4 @@
 Examples coming soon:
 
  - [docker-compose](docker-compose): Simple setup to run a scheduler and two clusters with local docker images
- - Kubernetes in Docker (kind)
+ - [kind](kind): Kubernetes in Docker (kind)

docs/examples/docker-compose/docker-compose-demo.yaml

@@ -7,7 +7,7 @@ services:
     container_name: scheduler
     image: ghcr.io/converged-computing/rainbow-scheduler:latest
     entrypoint: rainbow-scheduler
-    command: --address :8080 --name rainbow --secret peanutbuttajellay
+    command: --host :8080 --name rainbow --secret peanutbuttajellay
     volumes:
       - ./data:/data
     ports:

docs/examples/docker-compose/docker-compose.yaml

@@ -7,7 +7,7 @@ services:
     container_name: scheduler
     image: ghcr.io/converged-computing/rainbow-scheduler:latest
     entrypoint: rainbow-scheduler
-    command: --address :8080 --name rainbow --secret peanutbuttajellay
+    command: --host :8080 --name rainbow --secret peanutbuttajellay
     volumes:
       - ./data:/data
     ports:

docs/examples/kind/README.md

@@ -0,0 +1,198 @@
+# Kubernetes in Docker (kind)
+
+This example shows using the [rainbow docker images](https://github.com/orgs/converged-computing/packages?repo_name=rainbow) locally via kind, which is Kubernetes in docker. These same manifests (the YAML files) will likely work in production Kubernetes as well.
+
+## Usage
+
+## 1. Create Cluster
+
+This cluster is going to allow us to create ingress.
+
+```bash
+kind create cluster --config ./kind-config.yaml
+```
+
+## 2. Load Images
+
+This step is optional, but if you want to load your images in before creating Kubernetes objects, you can.
+This mostly helps if you have a local image (otherwise it will pull from the remote registry directory)
+
+```bash
+kind load docker-image ghcr.io/converged-computing/rainbow-flux:latest
+kind load docker-image ghcr.io/converged-computing/rainbow-scheduler:latest
+```
+
+## 3. Create Service and Ingress
+
+Let's next create the service. While we don't technically need this (communication happens within the network of pods) we anticipate some case when we will want to interact from outside of that space and thus show you how to set it up.
+
+```bash
+kubectl create -f ./service.yaml
+kubectl create -f ./ingress.yaml
+```
+
+## 4. Create Rainbow Scheduler
+
+Let's create the rainbow scheduler deployment.
+
+```bash
+kubectl create -f ./scheduler.yaml
+```
+
+And ensure it is running OK:
+
+```bash
+kubectl logs scheduler-798ddccf-pxfxx 
+```
+```console
+2024/02/14 20:53:37 creating 🌈️ server...
+2024/02/14 20:53:37 ✨️ creating rainbow.db...
+2024/02/14 20:53:37    rainbow.db file created
+2024/02/14 20:53:37    create cluster table...
+2024/02/14 20:53:37    cluster table created
+2024/02/14 20:53:37    create jobs table...
+2024/02/14 20:53:37    jobs table created
+2024/02/14 20:53:37 ⚠️ WARNING: global-token is set, use with caution.
+2024/02/14 20:53:37 starting scheduler server: rainbow v0.1.0-draft
+2024/02/14 20:53:37 server listening: [::]:8080
+```
+
+Importantly, we give the scheduler a predictable hostname. 
+
+```bash
+kubectl exec -it scheduler-798ddccf-pxfxx -- cat /etc/hosts
+```
+```console
+# Kubernetes-managed hosts file.
+127.0.0.1       localhost
+::1     localhost ip6-localhost ip6-loopback
+fe00::0 ip6-localnet
+fe00::0 ip6-mcastprefix
+fe00::1 ip6-allnodes
+fe00::2 ip6-allrouters
+10.244.0.5      scheduler.rainbow.default.svc.cluster.local     scheduler
+```
+
+## 5. Start Clusters
+
+We are going to cheat a bit and create multiple "clusters" (pods) via an indexed job.
+The names of the pods (hostnames) will correspond with the names of the clusters. We also
+are using a `--global-token` so a shared filesystem is not needed - the scheduler will assign
+the same token to all newly registered clusters. This is of course not intended for a production
+setup.
+
+```bash
+kubectl apply -f ./clusters.yaml
+```
+
+Four pods should be running now! You can watch the scheduler logs, and ultimately see logs of a cluster to see what is happening.
+
+```console
+👋️ Hello, I'm clusters-0!
+📜️ Registering clusters-0...
+token: "jellaytime"
+secret: "e2342db8-1c0b-40dd-9b42-e49e56480916"
+status: REGISTER_SUCCESS
+
+🥳️ All 3 clusters are registered.
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+Status: REQUEST_JOBS_SUCCESS
+Received 2 jobs for inspection!
+Accepting 2 jobs...
+[5, 1]
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'perfume']: ƒCYGUnZd
+Submit job ['echo', 'hello', 'from', 'clusters-2,', 'a', 'new', 'word', 'is', 'sn']: ƒCYraW3Z
+Ran job ƒCYGUnZd hello from clusters-1, a new word is perfume
+
+Ran job ƒCYraW3Z hello from clusters-2, a new word is sn
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+Status: REQUEST_JOBS_SUCCESS
+Received 2 jobs for inspection!
+Accepting 2 jobs...
+[16, 14]
+Submit job ['echo', 'hello', 'from', 'clusters-2,', 'a', 'new', 'word', 'is', 'world']: ƒH7VNJbZ
+Submit job ['echo', 'hello', 'from', 'clusters-0,', 'a', 'new', 'word', 'is', 'essex']: ƒH86x1Mq
+Ran job ƒH86x1Mq hello from clusters-0, a new word is essex
+
+Ran job ƒH7VNJbZ hello from clusters-2, a new word is world
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+Status: REQUEST_JOBS_SUCCESS
+Received 5 jobs for inspection!
+Accepting 5 jobs...
+[27, 23, 20, 29, 22]
+Submit job ['echo', 'hello', 'from', 'clusters-0,', 'a', 'new', 'word', 'is', 'resorts']: ƒMetzhd1
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'puzzles']: ƒMfY4Pfd
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'mean']: ƒMg9e6Ru
+Submit job ['echo', 'hello', 'from', 'clusters-0,', 'a', 'new', 'word', 'is', 'jamie']: ƒMgnhnUX
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'final']: ƒMhNoVxT
+Ran job ƒMhNoVxT hello from clusters-1, a new word is final
+
+Ran job ƒMg9e6Ru hello from clusters-1, a new word is mean
+
+Ran job ƒMgnhnUX hello from clusters-0, a new word is jamie
+
+Ran job ƒMfY4Pfd hello from clusters-1, a new word is puzzles
+
+Ran job ƒMetzhd1 hello from clusters-0, a new word is resorts
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+Status: REQUEST_JOBS_SUCCESS
+Received 2 jobs for inspection!
+Accepting 2 jobs...
+[33, 35]
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'goal']: ƒSDXnWB1
+Submit job ['echo', 'hello', 'from', 'clusters-1,', 'a', 'new', 'word', 'is', 'during']: ƒSE9NCwH
+Ran job ƒSE9NCwH hello from clusters-1, a new word is during
+
+Ran job ƒSDXnWB1 hello from clusters-1, a new word is goal
+
+status: SUBMIT_SUCCESS
+
+status: SUBMIT_SUCCESS
+
+Status: REQUEST_JOBS_SUCCESS
+Received 1 jobs for inspection!
+Accepting 1 jobs...
+[41]
+Submit job ['echo', 'hello', 'from', 'clusters-2,', 'a', 'new', 'word', 'is', 'manufacture']: ƒWjQTeHm
+Ran job ƒWjQTeHm hello from clusters-2, a new word is manufacture
+
+💤️ Cluster clusters-0 is finished! Shutting down.
+```
+
+
+And that's it! Clean up when you are done:
+
+```bash
+kind delete cluster
+```

docs/examples/kind/clusters.yaml

@@ -0,0 +1,24 @@
+apiVersion: batch/v1
+kind: Job
+metadata:
+  name: clusters
+spec:
+  completions: 3
+  parallelism: 3
+  completionMode: Indexed
+  template:
+    spec:
+      restartPolicy: Never
+      containers:
+      - name: cluster
+        image: ghcr.io/converged-computing/rainbow-flux:latest
+        command: ["flux"]
+
+        # Note that --host defaults to scheduler.rainbow.default.svc.cluster.local:8080
+        args: ["start", "python3",
+               "/code/docs/examples/kind/scripts/run-demo.py",
+               "--peer", "clusters-0",
+               "--peer", "clusters-1",
+               "--peer", "clusters-2",
+               "--iters", "5"]
+        imagePullPolicy: Never

docs/examples/kind/ingress.yaml

@@ -0,0 +1,17 @@
+apiVersion: networking.k8s.io/v1
+kind: Ingress
+metadata:
+  name: rainbow
+spec:
+  rules:
+  - host: localhost
+    http:
+      paths:
+      - path: /
+        pathType: Prefix
+        backend:
+          service:
+            name: rainbow
+            port: 
+              # TODO look at what wfmanager is doing with mlserver
+              number: 8080

docs/examples/kind/kind-config.yaml

@@ -0,0 +1,20 @@
+kind: Cluster
+apiVersion: kind.x-k8s.io/v1alpha4
+nodes:
+- role: control-plane
+  kubeadmConfigPatches:
+  - |
+    kind: InitConfiguration
+    nodeRegistration:
+      kubeletExtraArgs:
+        node-labels: "ingress-ready=true"
+  extraPortMappings:
+  - containerPort: 8080
+    hostPort: 8080
+    protocol: TCP
+  - containerPort: 80
+    hostPort: 80
+    protocol: TCP
+  - containerPort: 443
+    hostPort: 443
+    protocol: TCP

docs/examples/kind/scheduler.yaml

@@ -0,0 +1,26 @@
+apiVersion: apps/v1
+kind: Deployment
+metadata:
+  name: scheduler
+spec:
+  selector:
+    matchLabels:
+      app: rainbow
+  replicas: 1
+  template:
+    metadata:
+      labels:
+        # Matches the headless service
+        app: rainbow
+    spec:
+      subdomain: rainbow
+      hostname: scheduler
+      containers:
+      - name: scheduler
+        image: ghcr.io/converged-computing/rainbow-scheduler:latest
+
+        # Note that we are setting a global token (not recommended)! So that 
+        # we don't need a shared filesystem.
+        command: ["rainbow-scheduler"]
+        args: ["--host", ":8080", "--name", "rainbow", "--secret", "peanutbutta", "--global-token", "jellaytime"]
+        imagePullPolicy: Never