[core][3/3] Use the new standalone runtime env http server. (ray-proj…

…ect#37585) Rewrites agent_manager.cc. Removed its ability to do agent registration (no longer needs registration) and proxying runtime env agent (moved to the runtime_env_agent_client.cc). It will only do agent starting but we will have 2 instances in node_manager starting a dashboard agent and a runtime env agent. Deletes the runtime env agent python code from dashboard agent. Deletes the agent registration grpc interface, and the runtime env agent interface. Starts the standalone runtime env http server in services.py. Adds the extra port for the server everywhere: in services.py, node.py and gcs.proto. added 1 more port to Node info: runtime_env_agent_port. Intended to be used with raylet_address, but in some cases (1 test IIRC) we don't have one and it'll be used with node_address updated all related tests. Most tests used to use dashboard agent's port, now they use runtime env agent's port. Signed-off-by: NripeshN <[email protected]>
NripeshN · Aug 15, 2023 · e100c80 · e100c80
1 parent 44396f0
commit e100c80
Show file tree

Hide file tree

Showing 47 changed files with 673 additions and 1,413 deletions.
diff --git a/BUILD.bazel b/BUILD.bazel
@@ -255,58 +255,6 @@ cc_library(
  ],
 )
 
-# Agent manager.
-cc_grpc_library(
- name = "agent_manager_cc_grpc",
- srcs = ["//src/ray/protobuf:agent_manager_proto"],
- grpc_only = True,
- deps = ["//src/ray/protobuf:agent_manager_cc_proto"],
-)
-
-cc_library(
- name = "agent_manager_rpc",
- hdrs = glob([
- "src/ray/rpc/agent_manager/*.h",
- ]),
- copts = COPTS,
- strip_include_prefix = "src",
- deps = [
- ":agent_manager_cc_grpc",
- ":grpc_common_lib",
- ":ray_common",
- "@boost//:asio",
- "@com_github_grpc_grpc//:grpc++",
- ],
-)
-
-# runtime env.
-cc_grpc_library(
- name = "runtime_env_cc_grpc",
- srcs = ["//src/ray/protobuf:runtime_env_agent_proto"],
- grpc_only = True,
- deps = [
- "//src/ray/protobuf:agent_manager_cc_proto",
- "//src/ray/protobuf:common_cc_proto",
- "//src/ray/protobuf:runtime_env_agent_cc_proto",
- ],
-)
-
-cc_library(
- name = "runtime_env_rpc",
- hdrs = glob([
- "src/ray/rpc/runtime_env/*.h",
- ]),
- copts = COPTS,
- strip_include_prefix = "src",
- deps = [
- ":grpc_common_lib",
- ":ray_common",
- ":runtime_env_cc_grpc",
- "@boost//:asio",
- "@com_github_grpc_grpc//:grpc++",
- ],
-)
-
 # pubsub.
 cc_grpc_library(
  name = "pubsub_cc_grpc",
@@ -581,7 +529,6 @@ cc_library(
  copts = COPTS,
  strip_include_prefix = "src",
  deps = [
- ":agent_manager_rpc",
  ":autoscaler_rpc",
  ":gcs",
  ":gcs_pub_sub_lib",
@@ -595,6 +542,7 @@ cc_library(
  ":raylet_client_lib",
  ":scheduler",
  ":worker_rpc",
+ "//src/ray/protobuf:agent_manager_cc_proto",
  "@boost//:bimap",
  "@com_github_grpc_grpc//src/proto/grpc/health/v1:health_proto",
  "@com_google_absl//absl/container:btree",
@@ -774,7 +722,6 @@ cc_library(
  strip_include_prefix = "src",
  visibility = ["//visibility:public"],
  deps = [
- ":agent_manager_rpc",
  ":gcs",
  ":gcs_client_lib",
  ":node_manager_fbs",
@@ -783,7 +730,6 @@ cc_library(
  ":plasma_client",
  ":pubsub_lib",
  ":ray_common",
- ":runtime_env_rpc",
  ":scheduler",
  ":stats_lib",
  ":worker_rpc",
@@ -825,7 +771,6 @@ cc_library(
  strip_include_prefix = "src",
  visibility = ["//visibility:public"],
  deps = [
- ":agent_manager_rpc",
  ":node_manager_fbs",
  ":node_manager_rpc",
  ":ray_common",
@@ -2480,12 +2425,12 @@ cc_library(
  copts = COPTS,
  strip_include_prefix = "src",
  deps = [
- ":agent_manager_rpc",
  ":hiredis",
  ":node_manager_fbs",
  ":node_manager_rpc",
  ":ray_common",
  ":stats_lib",
+ "//src/ray/protobuf:agent_manager_cc_proto",
  "//src/ray/protobuf:gcs_cc_proto",
  "//src/ray/protobuf:gcs_service_cc_proto",
  "//src/ray/util",

diff --git a/dashboard/agent.py b/dashboard/agent.py
@@ -23,7 +23,6 @@
  setup_component_logger,
  configure_log_file,
 )
-from ray.core.generated import agent_manager_pb2, agent_manager_pb2_grpc
 from ray.experimental.internal_kv import (
  _initialize_internal_kv,
  _internal_kv_initialized,
@@ -84,7 +83,6 @@ def __init__(
  log_dir: str,
  temp_dir: str,
  session_dir: str,
- runtime_env_dir: str,
  logging_params: dict,
  agent_id: int,
  session_name: str,
@@ -99,7 +97,6 @@ def __init__(
 
  self.temp_dir = temp_dir
  self.session_dir = session_dir
- self.runtime_env_dir = runtime_env_dir
  self.log_dir = log_dir
  self.dashboard_agent_port = dashboard_agent_port
  self.metrics_export_port = metrics_export_port
@@ -321,19 +318,6 @@ async def _check_parent():
  namespace=ray_constants.KV_NAMESPACE_DASHBOARD,
  )
 
- # Register agent to agent manager.
- raylet_stub = agent_manager_pb2_grpc.AgentManagerServiceStub(
- self.aiogrpc_raylet_channel
- )
-
- await raylet_stub.RegisterAgent(
- agent_manager_pb2.RegisterAgentRequest(
- agent_id=self.agent_id,
- agent_port=self.grpc_port,
- agent_ip_address=self.ip,
- )
- )
-
  tasks = [m.run(self.server) for m in modules]
  if sys.platform not in ["win32", "cygwin"]:
  tasks.append(check_parent_task)
@@ -466,13 +450,7 @@ def open_capture_files(log_dir):
  default=None,
  help="Specify the path of this session.",
  )
- parser.add_argument(
- "--runtime-env-dir",
- required=True,
- type=str,
- default=None,
- help="Specify the path of the resource directory used by runtime_env.",
- )
+
  parser.add_argument(
  "--minimal",
  action="store_true",
@@ -530,7 +508,6 @@ def open_capture_files(log_dir):
  args.minimal,
  temp_dir=args.temp_dir,
  session_dir=args.session_dir,
- runtime_env_dir=args.runtime_env_dir,
  log_dir=args.log_dir,
  metrics_export_port=args.metrics_export_port,
  node_manager_port=args.node_manager_port,

diff --git a/dashboard/consts.py b/dashboard/consts.py
@@ -49,11 +49,6 @@
 # Default value for datacenter (the default value in protobuf)
 DEFAULT_LANGUAGE = "PYTHON"
 DEFAULT_JOB_ID = "ffff"
-# Cache TTL for bad runtime env. After this time, delete the cache and retry to create
-# runtime env if needed.
-BAD_RUNTIME_ENV_CACHE_TTL_SECONDS = env_integer(
- "BAD_RUNTIME_ENV_CACHE_TTL_SECONDS", 60 * 10
-)
 # Hook that is invoked on the dashboard `/api/component_activities` endpoint.
 # Environment variable stored here should be a callable that does not
 # take any arguments and should return a dictionary mapping

diff --git a/dashboard/modules/log/log_manager.py b/dashboard/modules/log/log_manager.py
@@ -126,7 +126,7 @@ async def stream_logs(
  yield streamed_log.data
 
  def _verify_node_registered(self, node_id: str):
- if node_id not in self.client.get_all_registered_agent_ids():
+ if node_id not in self.client.get_all_registered_log_agent_ids():
  raise DataSourceUnavailable(
  f"Given node id {node_id} is not available. "
  "It's either the node is dead, or it is not registered. "

diff --git a/dashboard/modules/runtime_env/__init__.py b/dashboard/modules/runtime_env/__init__.py