Cli submit delete raycluster (#257)

* add: create cluster from yaml function

* add: submit and delete functions

* change: cluster_name to name in submit raycluster

* add: load_auth in delete function

* update: make get_cluster function use new config

* test: unit tests for submit and delete raycluster commands

* change: format slightly on submit/delete commands

* Add: context for current namespace and .codeflare path

* fix: remove load_auth in functions so it doesn't run twice

* Add: help messages for submit and delete functions

* cleanup

* remove: remove get_namespace every function call

* fix: fix tests

* change: make namespace default to 'default'
and change test slightly

* refactor: remove unused imports

Author: carsonmh

Diff for: src/codeflare_sdk/cli/codeflare_cli.py

@@ -1,13 +1,21 @@
 import click
-import sys
 import os
 
+from codeflare_sdk.cli.cli_utils import load_auth
+
 cmd_folder = os.path.abspath(os.path.join(os.path.dirname(__file__), "commands"))
 
 
 class CodeflareContext:
-    def __init__(self, codeflare_path):
-        self.codeflare_path = codeflare_path
+    def __init__(self):
+        self.codeflare_path = _initialize_codeflare_folder()
+
+
+def _initialize_codeflare_folder():
+    codeflare_folder = os.path.expanduser("~/.codeflare")
+    if not os.path.exists(codeflare_folder):
+        os.makedirs(codeflare_folder)
+    return codeflare_folder
 
 
 class CodeflareCLI(click.MultiCommand):
@@ -31,18 +39,11 @@ def get_command(self, ctx, name):
             return
 
 
-def initialize_cli(ctx):
-    # Make .codeflare folder
-    codeflare_folder = os.path.expanduser("~/.codeflare")
-    if not os.path.exists(codeflare_folder):
-        os.makedirs(codeflare_folder)
-    ctx.obj = CodeflareContext(codeflare_folder)
-
-
 @click.command(cls=CodeflareCLI)
 @click.pass_context
 def cli(ctx):
-    initialize_cli(ctx)  # Ran on every command
+    load_auth()
+    ctx.obj = CodeflareContext()  # Ran on every command
     pass
 
 

Diff for: src/codeflare_sdk/cli/commands/delete.py

@@ -0,0 +1,23 @@
+import click
+
+from codeflare_sdk.cluster.cluster import get_cluster
+
+
+@click.group()
+def cli():
+    """
+    Delete a specified resource from the Kubernetes cluster
+    """
+    pass
+
+
+@cli.command()
+@click.argument("name", type=str)
+@click.option("--namespace", type=str, default="default")
+def raycluster(name, namespace):
+    """
+    Delete a specified RayCluster from the Kubernetes cluster
+    """
+    cluster = get_cluster(name, namespace)
+    cluster.down()
+    click.echo(f"Cluster deleted successfully")

Diff for: src/codeflare_sdk/cli/commands/submit.py

@@ -0,0 +1,32 @@
+import click
+
+from codeflare_sdk.cluster.cluster import Cluster
+
+
+@click.group()
+def cli():
+    """
+    Submit a defined resource to the Kubernetes cluster
+    """
+    pass
+
+
+@cli.command()
+@click.argument("name", type=str)
+@click.option("--wait", is_flag=True)
+def raycluster(name, wait):
+    """
+    Submit a defined RayCluster to the Kubernetes cluster
+    """
+    cluster = Cluster.from_definition_yaml(name + ".yaml")
+    if not cluster:
+        click.echo(
+            "Error submitting RayCluster. Make sure the RayCluster is defined before submitting it"
+        )
+        return
+    if not wait:
+        cluster.up()
+        click.echo("Cluster submitted successfully")
+        return
+    cluster.up()
+    cluster.wait_ready()

Diff for: src/codeflare_sdk/cluster/cluster.py

@@ -307,7 +307,8 @@ def torchx_config(
     def from_k8_cluster_object(rc):
         machine_types = (
             rc["metadata"]["labels"]["orderedinstance"].split("_")
-            if "orderedinstance" in rc["metadata"]["labels"]
+            if "labels" in rc["metadata"]
+            and "orderedinstance" in rc["metadata"]["labels"]
             else []
         )
         local_interactive = (
@@ -347,6 +348,58 @@ def from_k8_cluster_object(rc):
         )
         return Cluster(cluster_config)
 
+    def from_definition_yaml(yaml_path):
+        try:
+            with open(yaml_path) as yaml_file:
+                rc = yaml.load(yaml_file, Loader=yaml.FullLoader)
+                machine_types = (
+                    rc["metadata"]["labels"]["orderedinstance"].split("_")
+                    if "labels" in rc["metadata"]
+                    and "orderedinstance" in rc["metadata"]["labels"]
+                    else []
+                )
+                worker_group_specs = rc["spec"]["resources"]["GenericItems"][0][
+                    "generictemplate"
+                ]["spec"]["workerGroupSpecs"][0]
+                local_interactive = (
+                    "volumeMounts"
+                    in worker_group_specs["template"]["spec"]["containers"][0]
+                )
+                cluster_config = ClusterConfiguration(
+                    name=rc["metadata"]["name"],
+                    namespace=rc["metadata"]["namespace"],
+                    machine_types=machine_types,
+                    min_worker=worker_group_specs["minReplicas"],
+                    max_worker=worker_group_specs["maxReplicas"],
+                    min_cpus=worker_group_specs["template"]["spec"]["containers"][0][
+                        "resources"
+                    ]["requests"]["cpu"],
+                    max_cpus=worker_group_specs["template"]["spec"]["containers"][0][
+                        "resources"
+                    ]["limits"]["cpu"],
+                    min_memory=int(
+                        worker_group_specs["template"]["spec"]["containers"][0][
+                            "resources"
+                        ]["requests"]["memory"][:-1]
+                    ),
+                    max_memory=int(
+                        worker_group_specs["template"]["spec"]["containers"][0][
+                            "resources"
+                        ]["limits"]["memory"][:-1]
+                    ),
+                    gpu=worker_group_specs["template"]["spec"]["containers"][0][
+                        "resources"
+                    ]["requests"]["nvidia.com/gpu"],
+                    instascale=True if machine_types else False,
+                    image=worker_group_specs["template"]["spec"]["containers"][0][
+                        "image"
+                    ],
+                    local_interactive=local_interactive,
+                )
+                return Cluster(cluster_config)
+        except IOError:
+            return None
+
     def local_client_url(self):
         if self.config.local_interactive == True:
             ingress_domain = _get_ingress_domain()

Diff for: tests/cli-test-case.yaml

@@ -4,7 +4,7 @@ metadata:
   labels:
     orderedinstance: cpu.small_gpu.large
   name: cli-test-cluster
-  namespace: ns
+  namespace: default
 spec:
   priority: 9
   resources:
@@ -36,7 +36,7 @@ spec:
             appwrapper.mcad.ibm.com: cli-test-cluster
             controller-tools.k8s.io: '1.0'
           name: cli-test-cluster
-          namespace: ns
+          namespace: default
         spec:
           autoscalerOptions:
             idleTimeoutSeconds: 60
@@ -184,7 +184,7 @@ spec:
           labels:
             odh-ray-cluster-service: cli-test-cluster-head-svc
           name: ray-dashboard-cli-test-cluster
-          namespace: ns
+          namespace: default
         spec:
           port:
             targetPort: dashboard

Diff for: tests/unit_test.py

@@ -87,12 +87,13 @@ def test_cli_working():
     assert result.exit_code == 0
 
 
-def test_cluster_definition_cli():
+def test_cluster_definition_cli(mocker):
+    mocker.patch.object(client, "ApiClient")
     runner = CliRunner()
     define_cluster_command = """
                         define raycluster
                         --name=cli-test-cluster
-                        --namespace=ns
+                        --namespace=default
                         --min_worker=1
                         --max_worker=2
                         --min_cpus=3
@@ -105,7 +106,10 @@ def test_cluster_definition_cli():
                         --image_pull_secrets='["cli-test-pull-secret"]'
                         """
     result = runner.invoke(cli, define_cluster_command)
-    assert result.output == "Written to: cli-test-cluster.yaml\n"
+    assert (
+        result.output
+        == "No authentication found, trying default kubeconfig\nWritten to: cli-test-cluster.yaml\n"
+    )
     assert filecmp.cmp(
         "cli-test-cluster.yaml", f"{parent}/tests/cli-test-case.yaml", shallow=True
     )
@@ -120,7 +124,10 @@ def test_login_cli(mocker):
                         --token=testtoken
                         """
     login_result = runner.invoke(cli, k8s_login_command)
-    assert login_result.output == "Logged into 'testserver:6443'\n"
+    assert (
+        login_result.output
+        == "No authentication found, trying default kubeconfig\nLogged into 'testserver:6443'\n"
+    )
     try:
         auth_file_path = os.path.expanduser("~/.codeflare/auth")
         with open(auth_file_path, "rb") as file:
@@ -170,6 +177,37 @@ def test_load_auth():
     assert sdk_auth.api_client is not None
 
 
+def test_cluster_submission_cli(mocker):
+    mocker.patch.object(client, "ApiClient")
+    runner = CliRunner()
+    submit_cluster_command = """
+                            submit raycluster
+                            cli-test-cluster
+                            """
+    result = runner.invoke(cli, submit_cluster_command)
+
+    assert result.exit_code == 0
+    assert "Cluster submitted successfully" in result.output
+
+
+def test_cluster_deletion_cli(mocker):
+    mocker.patch.object(client, "ApiClient")
+    mocker.patch("kubernetes.config.load_kube_config", return_value="ignore")
+    mocker.patch(
+        "kubernetes.client.CustomObjectsApi.list_namespaced_custom_object",
+        side_effect=get_ray_obj,
+    )
+    runner = CliRunner()
+    delete_cluster_command = """
+                            delete raycluster
+                            quicktest
+                            """
+    result = runner.invoke(cli, delete_cluster_command)
+
+    assert result.exit_code == 0
+    assert "Cluster deleted successfully" in result.output
+
+
 # For mocking openshift client results
 fake_res = openshift.Result("fake")