Merge branch 'master' into offline-rl-handle-duplicates-in-buffer

doc/source/cluster/metrics.md

@@ -22,6 +22,12 @@ For a quick demo, you can run Prometheus locally on your machine. Follow the qui
 
 ### Quickstart: Running Prometheus locally
 
+```{admonition} Note
+:class: note
+If you need to change the root temporary directory by using "--temp-dir" in your Ray 
+cluster setup, follow these [manual steps](#optional-manual-running-prometheus-locally) to set up Prometheus locally.
+```
+
 Run the following command to download and start Prometheus locally with a configuration that scrapes metrics from a local Ray Cluster.
 
 ```bash
@@ -76,7 +82,7 @@ tar xvfz prometheus-*.tar.gz
 cd prometheus-*
 ```
 
-Ray provides a Prometheus config that works out of the box. After running Ray, you can find the config at `/tmp/ray/session_latest/metrics/prometheus/prometheus.yml`.
+Ray provides a Prometheus config that works out of the box. After running Ray, you can find the config at `/tmp/ray/session_latest/metrics/prometheus/prometheus.yml`. If you specify the `--temp-dir={your_temp_path}` when starting the Ray cluster, the config file is at `{your_temp_path}/session_latest/metrics/prometheus/prometheus.yml`
 
 ```yaml
 global:
@@ -88,13 +94,17 @@ scrape_configs:
 - job_name: 'ray'
   file_sd_configs:
   - files:
-    - '/tmp/ray/prom_metrics_service_discovery.json'
+    - '/tmp/ray/prom_metrics_service_discovery.json' # or '${your_temp_path}/prom_metrics_service_discovery.json' if --temp-dir is specified
 ```
 
 Next, start Prometheus:
 
 ```shell
+# With default settings
 ./prometheus --config.file=/tmp/ray/session_latest/metrics/prometheus/prometheus.yml
+
+# With specified --temp-dir
+./prometheus --config.file={your_temp_path}/session_latest/metrics/prometheus/prometheus.yml
 ```
 ```{admonition} Note
 :class: note

doc/source/serve/api/index.md

@@ -373,6 +373,7 @@ Content-Type: application/json
    :toctree: doc/
 
    schema.ServeInstanceDetails
+   schema.APIType
    schema.ApplicationDetails
    schema.DeploymentDetails
    schema.ReplicaDetails

python/ray/dashboard/BUILD

@@ -23,8 +23,8 @@ py_test_run_all_subdirectory(
         "client/node_modules/**",
         "modules/job/tests/test_cli_integration.py",
         "modules/job/tests/test_http_job_server.py",
-        "modules/node/tests/test_node.py", 
-        "tests/test_dashboard.py", 
+        "modules/node/tests/test_node.py",
+        "tests/test_dashboard.py",
         "tests/test_state_head.py",
         "modules/serve/tests/test_serve_dashboard.py",
     ],
@@ -100,7 +100,7 @@ py_test(
 py_test(
     name = "test_serve_dashboard",
     size = "enormous",
-    srcs = ["modules/serve/tests/test_serve_dashboard.py"],
+    srcs = ["modules/serve/tests/test_serve_dashboard.py", "modules/serve/tests/deploy_imperative_serve_apps.py"],
     tags = ["team:serve"],
     deps = [":conftest"],
 )

python/ray/dashboard/modules/metrics/grafana_dashboard_factory.py

@@ -19,9 +19,6 @@
     serve_deployment_dashboard_config,
 )
 
-METRICS_INPUT_ROOT = os.path.join(os.path.dirname(__file__), "export")
-GRAFANA_CONFIG_INPUT_PATH = os.path.join(METRICS_INPUT_ROOT, "grafana")
-
 GRAFANA_DASHBOARD_UID_OVERRIDE_ENV_VAR_TEMPLATE = "RAY_GRAFANA_{name}_DASHBOARD_UID"
 GRAFANA_DASHBOARD_GLOBAL_FILTERS_OVERRIDE_ENV_VAR_TEMPLATE = (
     "RAY_GRAFANA_{name}_DASHBOARD_GLOBAL_FILTERS"

python/ray/dashboard/modules/metrics/install_and_start_prometheus.py

@@ -74,12 +74,13 @@ def install_prometheus(file_path):
 
 def start_prometheus(prometheus_dir):
 
-    # Currently, Ray never modifies this config file, so we can just use the
-    # hardcoded path. (It just copies it to a more user-friendly location, in
-    # MetricsHead._create_default_prometheus_configs.)
-    # However, if in the future Ray ever modifies this file at runtime, we'll
-    # need to use the user-friendly location instead, and reload the config
-    # file after it's updated by Ray.
+    # The function assumes the Ray cluster to be monitored by Prometheus uses the
+    # default configuration with "/tmp/ray" as the default root temporary directory.
+    #
+    # This is to support the `ray metrics launch-prometheus` command, when a ray cluster
+    # is not started yet and we have no way to get a `--temp-dir` anywhere. So we choose
+    # to use a hardcoded default value.
+
     config_file = Path(PROMETHEUS_CONFIG_INPUT_PATH)
 
     if not config_file.exists():

python/ray/dashboard/modules/metrics/metrics_head.py

@@ -11,24 +11,24 @@
 import ray.dashboard.optional_utils as dashboard_optional_utils
 import ray.dashboard.utils as dashboard_utils
 from ray._private.async_utils import enable_monitor_loop_lag
-from ray._private.ray_constants import env_integer
-from ray._private.utils import get_or_create_event_loop
-from ray.dashboard.consts import (
-    AVAILABLE_COMPONENT_NAMES_FOR_METRICS,
-    METRICS_INPUT_ROOT,
-    PROMETHEUS_CONFIG_INPUT_PATH,
+from ray._private.ray_constants import (
+    PROMETHEUS_SERVICE_DISCOVERY_FILE,
+    SESSION_LATEST,
+    env_integer,
 )
+from ray._private.utils import get_or_create_event_loop
+from ray.dashboard.consts import AVAILABLE_COMPONENT_NAMES_FOR_METRICS
 from ray.dashboard.modules.metrics.grafana_dashboard_factory import (
     generate_data_grafana_dashboard,
     generate_default_grafana_dashboard,
     generate_serve_deployment_grafana_dashboard,
     generate_serve_grafana_dashboard,
 )
-from ray.dashboard.modules.metrics.grafana_dashboard_provisioning_template import (
+from ray.dashboard.modules.metrics.templates import (
     DASHBOARD_PROVISIONING_TEMPLATE,
-)
-from ray.dashboard.modules.metrics.grafana_datasource_template import (
     GRAFANA_DATASOURCE_TEMPLATE,
+    GRAFANA_INI_TEMPLATE,
+    PROMETHEUS_YML_TEMPLATE,
 )
 
 import psutil
@@ -54,7 +54,6 @@
 GRAFANA_HOST_DISABLED_VALUE = "DISABLED"
 GRAFANA_IFRAME_HOST_ENV_VAR = "RAY_GRAFANA_IFRAME_HOST"
 GRAFANA_DASHBOARD_OUTPUT_DIR_ENV_VAR = "RAY_METRICS_GRAFANA_DASHBOARD_OUTPUT_DIR"
-GRAFANA_CONFIG_INPUT_PATH = os.path.join(METRICS_INPUT_ROOT, "grafana")
 GRAFANA_HEALTHCHECK_PATH = "api/health"
 
 
@@ -75,13 +74,22 @@ def __init__(self, dashboard_head):
             PROMETHEUS_HOST_ENV_VAR, DEFAULT_PROMETHEUS_HOST
         )
         default_metrics_root = os.path.join(self._dashboard_head.session_dir, "metrics")
+        session_latest_metrics_root = os.path.join(
+            self._dashboard_head.temp_dir, SESSION_LATEST, "metrics"
+        )
         self._metrics_root = os.environ.get(
             METRICS_OUTPUT_ROOT_ENV_VAR, default_metrics_root
         )
-        grafana_config_output_path = os.path.join(self._metrics_root, "grafana")
+        self._metrics_root_session_latest = os.environ.get(
+            METRICS_OUTPUT_ROOT_ENV_VAR, session_latest_metrics_root
+        )
+        self._grafana_config_output_path = os.path.join(self._metrics_root, "grafana")
+        self._grafana_session_latest_config_output_path = os.path.join(
+            self._metrics_root_session_latest, "grafana"
+        )
         self._grafana_dashboard_output_dir = os.environ.get(
             GRAFANA_DASHBOARD_OUTPUT_DIR_ENV_VAR,
-            os.path.join(grafana_config_output_path, "dashboards"),
+            os.path.join(self._grafana_config_output_path, "dashboards"),
         )
 
         self._prometheus_name = os.environ.get(
@@ -186,17 +194,34 @@ def _create_default_grafana_configs(self):
         """
         Creates the grafana configurations that are by default provided by Ray.
         """
-        grafana_config_output_path = os.path.join(self._metrics_root, "grafana")
-
-        # Copy default grafana configurations
-        if os.path.exists(grafana_config_output_path):
-            shutil.rmtree(grafana_config_output_path)
-        os.makedirs(os.path.dirname(grafana_config_output_path), exist_ok=True)
-        shutil.copytree(GRAFANA_CONFIG_INPUT_PATH, grafana_config_output_path)
+        # Create grafana configuration folder
+        if os.path.exists(self._grafana_config_output_path):
+            shutil.rmtree(self._grafana_config_output_path)
+        os.makedirs(self._grafana_config_output_path, exist_ok=True)
+
+        # Overwrite grafana's configuration file
+        grafana_provisioning_folder = os.path.join(
+            self._grafana_config_output_path, "provisioning"
+        )
+        grafana_prov_folder_with_latest_session = os.path.join(
+            self._grafana_session_latest_config_output_path, "provisioning"
+        )
+        with open(
+            os.path.join(
+                self._grafana_config_output_path,
+                "grafana.ini",
+            ),
+            "w",
+        ) as f:
+            f.write(
+                GRAFANA_INI_TEMPLATE.format(
+                    grafana_provisioning_folder=grafana_prov_folder_with_latest_session
+                )
+            )
 
         # Overwrite grafana's dashboard provisioning directory based on env var
         dashboard_provisioning_path = os.path.join(
-            grafana_config_output_path, "provisioning", "dashboards"
+            grafana_provisioning_folder, "dashboards"
         )
         os.makedirs(
             dashboard_provisioning_path,
@@ -219,9 +244,7 @@ def _create_default_grafana_configs(self):
         prometheus_host = os.environ.get(
             PROMETHEUS_HOST_ENV_VAR, DEFAULT_PROMETHEUS_HOST
         )
-        data_sources_path = os.path.join(
-            grafana_config_output_path, "provisioning", "datasources"
-        )
+        data_sources_path = os.path.join(grafana_provisioning_folder, "datasources")
         os.makedirs(
             data_sources_path,
             exist_ok=True,
@@ -297,14 +320,26 @@ def _create_default_prometheus_configs(self):
             self._metrics_root, "prometheus", "prometheus.yml"
         )
 
-        # Copy default prometheus configurations
+        # Generate the default prometheus configurations
         if os.path.exists(prometheus_config_output_path):
             os.remove(prometheus_config_output_path)
         os.makedirs(os.path.dirname(prometheus_config_output_path), exist_ok=True)
-        # Currently Ray directly copies this file without modifying it at runtime.
-        # If Ray ever modifies this file at runtime, please ensure start_prometheus
-        # in install_and_start_prometheus.py is updated to reload the config file.
-        shutil.copy(PROMETHEUS_CONFIG_INPUT_PATH, prometheus_config_output_path)
+
+        # This code generates the Prometheus config based on the custom temporary root
+        # path set by the user at Ray cluster start up (via --temp-dir). In contrast,
+        # start_prometheus in install_and_start_prometheus.py uses a hardcoded
+        # Prometheus config at PROMETHEUS_CONFIG_INPUT_PATH that always uses "/tmp/ray".
+        # Other than the root path, the config file generated here is identical to that
+        # hardcoded config file.
+        prom_discovery_file_path = os.path.join(
+            self._dashboard_head.temp_dir, PROMETHEUS_SERVICE_DISCOVERY_FILE
+        )
+        with open(prometheus_config_output_path, "w") as f:
+            f.write(
+                PROMETHEUS_YML_TEMPLATE.format(
+                    prom_metrics_service_discovery_file_path=prom_discovery_file_path
+                )
+            )
 
     @dashboard_utils.async_loop_forever(METRICS_RECORD_INTERVAL_S)
     async def record_dashboard_metrics(self):

python/ray/dashboard/modules/metrics/templates.py

@@ -0,0 +1,49 @@
+GRAFANA_INI_TEMPLATE = """
+[security]
+allow_embedding = true
+
+[auth.anonymous]
+enabled = true
+org_name = Main Org.
+org_role = Viewer
+
+[paths]
+provisioning = {grafana_provisioning_folder}
+"""
+
+DASHBOARD_PROVISIONING_TEMPLATE = """
+apiVersion: 1
+
+providers:
+  - name: Ray    # Default dashboards provided by OSS ray
+    folder: Ray
+    type: file
+    options:
+      path: {dashboard_output_folder}
+"""
+
+GRAFANA_DATASOURCE_TEMPLATE = """apiVersion: 1
+
+datasources:
+  - name: {prometheus_name}
+    url: {prometheus_host}
+    type: prometheus
+    isDefault: true
+    access: proxy
+"""
+
+PROMETHEUS_YML_TEMPLATE = """# my global config
+global:
+  scrape_interval: 10s # Set the scrape interval to every 10 seconds. Default is every \
+1 minute.
+  evaluation_interval: 10s # Evaluate rules every 10 seconds. The default is every 1 \
+minute.
+  # scrape_timeout is set to the global default (10s).
+
+scrape_configs:
+# Scrape from each ray node as defined in the service_discovery.json provided by ray.
+- job_name: 'ray'
+  file_sd_configs:
+  - files:
+    - '{prom_metrics_service_discovery_file_path}'
+"""

python/ray/dashboard/modules/serve/tests/deploy_imperative_serve_apps.py

@@ -0,0 +1,16 @@
+# Run in a subprocess by test_get_serve_instance_details_for_imperative_apps
+
+from ray import serve
+from ray.serve.tests.test_config_files import world
+
+
+def submit_imperative_apps() -> None:
+    serve.run(world.DagNode, name="app1", route_prefix="/apple")
+    print("Deployed app1")
+
+    serve.run(world.DagNode, name="app2", route_prefix="/banana")
+    print("Deployed app2")
+
+
+if __name__ == "__main__":
+    submit_imperative_apps()