indexify 0.3.16__py3-none-any.whl → 0.3.18__py3-none-any.whl

indexify/executor/grpc/state_reporter.py

@@ -1,37 +1,44 @@
 import asyncio
+import hashlib
+from socket import gethostname
 from typing import Any, Dict, List, Optional
 
 import grpc
 
-from indexify.proto.task_scheduler_pb2 import (
+from indexify.proto.executor_api_pb2 import (
     AllowedFunction,
+)
+from indexify.proto.executor_api_pb2 import ExecutorFlavor as ExecutorFlavorProto
+from indexify.proto.executor_api_pb2 import (
     ExecutorState,
     ExecutorStatus,
     FunctionExecutorDescription,
 )
-from indexify.proto.task_scheduler_pb2 import (
+from indexify.proto.executor_api_pb2 import (
     FunctionExecutorState as FunctionExecutorStateProto,
 )
-from indexify.proto.task_scheduler_pb2 import (
+from indexify.proto.executor_api_pb2 import (
     FunctionExecutorStatus as FunctionExecutorStatusProto,
 )
-from indexify.proto.task_scheduler_pb2 import (
+from indexify.proto.executor_api_pb2 import (
     GPUModel,
     GPUResources,
     HostResources,
     ReportExecutorStateRequest,
 )
-from indexify.proto.task_scheduler_pb2_grpc import (
-    TaskSchedulerServiceStub,
+from indexify.proto.executor_api_pb2_grpc import (
+    ExecutorAPIStub,
 )
 
 from ..api_objects import FunctionURI
+from ..executor_flavor import ExecutorFlavor
 from ..function_executor.function_executor_state import FunctionExecutorState
 from ..function_executor.function_executor_states_container import (
     FunctionExecutorStatesContainer,
 )
 from ..function_executor.function_executor_status import FunctionExecutorStatus
-from .channel_creator import ChannelCreator
+from ..runtime_probes import RuntimeProbes
+from .channel_manager import ChannelManager
 from .metrics.state_reporter import (
     metric_state_report_errors,
     metric_state_report_latency,
@@ -47,24 +54,32 @@ class ExecutorStateReporter:
     def __init__(
         self,
         executor_id: str,
+        flavor: ExecutorFlavor,
+        version: str,
+        labels: Dict[str, str],
         development_mode: bool,
         function_allowlist: Optional[List[FunctionURI]],
         function_executor_states: FunctionExecutorStatesContainer,
-        channel_creator: ChannelCreator,
+        channel_manager: ChannelManager,
         logger: Any,
     ):
         self._executor_id: str = executor_id
+        self._flavor: ExecutorFlavor = flavor
+        self._version: str = version
+        self._labels: Dict[str, str] = labels.copy()
         self._development_mode: bool = development_mode
+        self._hostname: str = gethostname()
         self._function_executor_states: FunctionExecutorStatesContainer = (
             function_executor_states
         )
-        self._channel_creator = channel_creator
+        self._channel_manager = channel_manager
         self._logger: Any = logger.bind(module=__name__)
         self._is_shutdown: bool = False
         self._executor_status: ExecutorStatus = ExecutorStatus.EXECUTOR_STATUS_UNKNOWN
         self._allowed_functions: List[AllowedFunction] = _to_grpc_allowed_functions(
             function_allowlist
         )
+        self._labels.update(_label_values_to_strings(RuntimeProbes().probe().labels))
 
     def update_executor_status(self, value: ExecutorStatus):
         self._executor_status = value
@@ -75,12 +90,16 @@ class ExecutorStateReporter:
         Never raises any exceptions.
         """
         while not self._is_shutdown:
-            async with await self._channel_creator.create() as server_channel:
+            async with await self._channel_manager.get_channel() as server_channel:
                 server_channel: grpc.aio.Channel
-                stub = TaskSchedulerServiceStub(server_channel)
+                stub = ExecutorAPIStub(server_channel)
                 while not self._is_shutdown:
                     try:
-                        await self._report_state(stub)
+                        # The periodic state reports serve as channel health monitoring requests
+                        # (same as TCP keep-alive). Channel Manager returns the same healthy channel
+                        # for all RPCs that we do from Executor to Server. So all the RPCs benefit
+                        # from this channel health monitoring.
+                        await self.report_state(stub)
                         await asyncio.sleep(_REPORTING_INTERVAL_SEC)
                     except Exception as e:
                         self._logger.error(
@@ -92,7 +111,11 @@ class ExecutorStateReporter:
 
         self._logger.info("State reporter shutdown")
 
-    async def _report_state(self, stub: TaskSchedulerServiceStub):
+    async def report_state(self, stub: ExecutorAPIStub):
+        """Reports the current state to the server represented by the supplied stub.
+
+        Raises exceptions on failure.
+        """
         with (
             metric_state_report_errors.count_exceptions(),
             metric_state_report_latency.time(),
@@ -101,11 +124,16 @@ class ExecutorStateReporter:
             state = ExecutorState(
                 executor_id=self._executor_id,
                 development_mode=self._development_mode,
-                executor_status=self._executor_status,
+                hostname=self._hostname,
+                flavor=_to_grpc_executor_flavor(self._flavor, self._logger),
+                version=self._version,
+                status=self._executor_status,
                 free_resources=await self._fetch_free_host_resources(),
                 allowed_functions=self._allowed_functions,
                 function_executor_states=await self._fetch_function_executor_states(),
+                labels=self._labels,
             )
+            state.state_hash = _state_hash(state)
 
             await stub.report_executor_state(
                 ReportExecutorStateRequest(executor_state=state),
@@ -197,3 +225,33 @@ def _to_grpc_function_executor_status(
         logger.error("Unexpected Function Executor status", status=status)
 
     return result
+
+
+_FLAVOR_MAPPING = {
+    ExecutorFlavor.OSS: ExecutorFlavorProto.EXECUTOR_FLAVOR_OSS,
+    ExecutorFlavor.PLATFORM: ExecutorFlavorProto.EXECUTOR_FLAVOR_PLATFORM,
+}
+
+
+def _to_grpc_executor_flavor(
+    flavor: ExecutorFlavor, logger: Any
+) -> ExecutorFlavorProto:
+    result: ExecutorFlavorProto = _FLAVOR_MAPPING.get(
+        flavor, ExecutorFlavorProto.EXECUTOR_FLAVOR_UNKNOWN
+    )
+
+    if result == ExecutorFlavorProto.EXECUTOR_FLAVOR_UNKNOWN:
+        logger.error("Unexpected Executor flavor", flavor=flavor)
+
+    return result
+
+
+def _label_values_to_strings(labels: Dict[str, Any]) -> Dict[str, str]:
+    return {k: str(v) for k, v in labels.items()}
+
+
+def _state_hash(state: ExecutorState) -> str:
+    serialized_state: bytes = state.SerializeToString(deterministic=True)
+    hasher = hashlib.sha256(usedforsecurity=False)
+    hasher.update(serialized_state)
+    return hasher.hexdigest()

indexify/executor/metrics/task_runner.py

@@ -23,6 +23,13 @@ metric_tasks_blocked_by_policy: prometheus_client.Gauge = prometheus_client.Gaug
     "tasks_blocked_by_policy",
     "Number of tasks that are ready for execution but are blocked according to the current policy (typically waiting for a free Function Executor)",
 )
+metric_tasks_blocked_by_policy_per_function_name: prometheus_client.Gauge = (
+    prometheus_client.Gauge(
+        "tasks_blocked_by_policy_per_function_name",
+        "Number of tasks that are ready for execution but are blocked according to the current policy (typically waiting for a free Function Executor)",
+        ["function_name"],
+    )
+)
 
 # Metrics for the stage when task is running.
 metric_task_runs: prometheus_client.Counter = prometheus_client.Counter(

indexify/executor/task_fetcher.py

@@ -1,6 +1,7 @@
 import json
 import time
-from typing import AsyncGenerator, List, Optional
+from socket import gethostname
+from typing import AsyncGenerator, Dict, List, Optional
 
 import structlog
 from httpx_sse import aconnect_sse
@@ -22,6 +23,7 @@ class TaskFetcher:
         self,
         executor_id: str,
         executor_version: str,
+        labels: Dict[str, str],
         function_allowlist: Optional[List[FunctionURI]],
         protocol: str,
         indexify_server_addr: str,
@@ -33,12 +35,15 @@ class TaskFetcher:
         self._logger = structlog.get_logger(module=__name__)
 
         probe_info: ProbeInfo = RuntimeProbes().probe()
+        all_labels = probe_info.labels.copy()
+        all_labels.update(labels)
+
         self._executor_metadata: ExecutorMetadata = ExecutorMetadata(
             id=executor_id,
             executor_version=executor_version,
-            addr="",
+            addr=gethostname(),
             function_allowlist=function_allowlist,
-            labels=probe_info.labels,
+            labels=all_labels,
         )
 
     async def run(self) -> AsyncGenerator[Task, None]:

indexify/executor/task_reporter.py

@@ -49,6 +49,7 @@ class TaskReporter:
     ):
         self._base_url = base_url
         self._executor_id = executor_id
+        self._is_shutdown = False
         # Use thread-safe sync client due to issues with async client.
         # Async client attempts to use connections it already closed.
         # See e.g. https://github.com/encode/httpx/issues/2337.
@@ -56,9 +57,25 @@ class TaskReporter:
         # results in not reusing established TCP connections to server.
         self._client = get_httpx_client(config_path, make_async=False)
 
+    async def shutdown(self):
+        """Shuts down the task reporter.
+
+        Task reporter stops reporting all task outcomes to the Server.
+        There are many task failures due to Executor shutdown. We give wrong
+        signals to Server if we report such failures.
+        """
+        self._is_shutdown = True
+
     async def report(self, output: TaskOutput, logger: Any):
         """Reports result of the supplied task."""
         logger = logger.bind(module=__name__)
+
+        if self._is_shutdown:
+            logger.warning(
+                "task reporter got shutdown, skipping task outcome reporting"
+            )
+            return
+
         task_result, output_files, output_summary = self._process_task_output(output)
         task_result_data = task_result.model_dump_json(exclude_none=True)
 

indexify/executor/task_runner.py

@@ -22,6 +22,7 @@ from .metrics.task_runner import (
     metric_task_run_platform_errors,
     metric_task_runs,
     metric_tasks_blocked_by_policy,
+    metric_tasks_blocked_by_policy_per_function_name,
     metric_tasks_running,
 )
 
@@ -55,6 +56,9 @@ class TaskRunner:
             with (
                 metric_task_policy_errors.count_exceptions(),
                 metric_tasks_blocked_by_policy.track_inprogress(),
+                metric_tasks_blocked_by_policy_per_function_name.labels(
+                    function_name=task_input.task.compute_fn
+                ).track_inprogress(),
                 metric_task_policy_latency.time(),
             ):
                 metric_task_policy_runs.inc()

indexify/proto/{task_scheduler.proto → executor_api.proto}

@@ -1,6 +1,8 @@
 syntax = "proto3";
 
-package task_scheduler_service;
+// Rename with caution. The package name is part of gRPC service name.
+// Existing clients won't find the service if the package name changes.
+package executor_api_pb;
 
 // ===== ReportExecutorState RPC =====
 
@@ -81,15 +83,26 @@ enum ExecutorStatus {
     EXECUTOR_STATUS_STOPPED = 5;
 }
 
+enum ExecutorFlavor {
+    EXECUTOR_FLAVOR_UNKNOWN = 0;
+    EXECUTOR_FLAVOR_OSS = 1;
+    EXECUTOR_FLAVOR_PLATFORM = 2;
+}
+
 message ExecutorState {
     optional string executor_id = 1;
     optional bool development_mode = 2;
-    optional ExecutorStatus executor_status = 3;
+    optional string hostname = 3;
+    optional ExecutorFlavor flavor = 4;
+    optional string version = 5;
+    optional ExecutorStatus status = 6;
     // Free resources available at the Executor.
-    optional HostResources free_resources = 4;
+    optional HostResources free_resources = 7;
     // Empty allowed_functions list means that any function can run on the Executor.
-    repeated AllowedFunction allowed_functions = 5;
-    repeated FunctionExecutorState function_executor_states = 6;
+    repeated AllowedFunction allowed_functions = 8;
+    repeated FunctionExecutorState function_executor_states = 9;
+    map<string, string> labels = 10;
+    optional string state_hash = 11;
 }
 
 // A message sent by Executor to report its up to date state to Server.
@@ -136,7 +149,11 @@ message DesiredExecutorState {
 
 // Internal API for scheduling and running tasks on Executors. Executors are acting as clients of this API.
 // Server is responsible for scheduling tasks on Executors and Executors are responsible for running the tasks.
-service TaskSchedulerService {
+//
+// Rename with caution. Existing clients won't find the service if the service name changes. A HTTP2 ingress proxy
+// might use the service name in it HTTP2 path based routing rules. See how gRPC uses service names in its HTTP2 paths
+// at https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md.
+service ExecutorAPI {
     // Called by Executor every 5 seconds to report that it's still alive and provide its current state.
     //
     // Missing 3 reports will result in the Executor being deregistered by Server.

indexify/proto/executor_api_pb2.py

@@ -0,0 +1,70 @@
+# -*- coding: utf-8 -*-
+# Generated by the protocol buffer compiler.  DO NOT EDIT!
+# NO CHECKED-IN PROTOBUF GENCODE
+# source: indexify/proto/executor_api.proto
+# Protobuf Python Version: 5.29.0
+"""Generated protocol buffer code."""
+from google.protobuf import descriptor as _descriptor
+from google.protobuf import descriptor_pool as _descriptor_pool
+from google.protobuf import runtime_version as _runtime_version
+from google.protobuf import symbol_database as _symbol_database
+from google.protobuf.internal import builder as _builder
+
+_runtime_version.ValidateProtobufRuntimeVersion(
+    _runtime_version.Domain.PUBLIC, 5, 29, 0, "", "indexify/proto/executor_api.proto"
+)
+# @@protoc_insertion_point(imports)
+
+_sym_db = _symbol_database.Default()
+
+
+DESCRIPTOR = _descriptor_pool.Default().AddSerializedFile(
+    b'\n!indexify/proto/executor_api.proto\x12\x0f\x65xecutor_api_pb"e\n\x0cGPUResources\x12\x12\n\x05\x63ount\x18\x01 \x01(\rH\x00\x88\x01\x01\x12-\n\x05model\x18\x02 \x01(\x0e\x32\x19.executor_api_pb.GPUModelH\x01\x88\x01\x01\x42\x08\n\x06_countB\x08\n\x06_model"\xc2\x01\n\rHostResources\x12\x16\n\tcpu_count\x18\x01 \x01(\rH\x00\x88\x01\x01\x12\x19\n\x0cmemory_bytes\x18\x02 \x01(\x04H\x01\x88\x01\x01\x12\x17\n\ndisk_bytes\x18\x03 \x01(\x04H\x02\x88\x01\x01\x12/\n\x03gpu\x18\x04 \x01(\x0b\x32\x1d.executor_api_pb.GPUResourcesH\x03\x88\x01\x01\x42\x0c\n\n_cpu_countB\x0f\n\r_memory_bytesB\r\n\x0b_disk_bytesB\x06\n\x04_gpu"\xbb\x01\n\x0f\x41llowedFunction\x12\x16\n\tnamespace\x18\x01 \x01(\tH\x00\x88\x01\x01\x12\x17\n\ngraph_name\x18\x02 \x01(\tH\x01\x88\x01\x01\x12\x1a\n\rfunction_name\x18\x03 \x01(\tH\x02\x88\x01\x01\x12\x1a\n\rgraph_version\x18\x04 \x01(\tH\x03\x88\x01\x01\x42\x0c\n\n_namespaceB\r\n\x0b_graph_nameB\x10\n\x0e_function_nameB\x10\n\x0e_graph_version"\xed\x02\n\x1b\x46unctionExecutorDescription\x12\x0f\n\x02id\x18\x01 \x01(\tH\x00\x88\x01\x01\x12\x16\n\tnamespace\x18\x02 \x01(\tH\x01\x88\x01\x01\x12\x17\n\ngraph_name\x18\x03 \x01(\tH\x02\x88\x01\x01\x12\x1a\n\rgraph_version\x18\x04 \x01(\tH\x03\x88\x01\x01\x12\x1a\n\rfunction_name\x18\x05 \x01(\tH\x04\x88\x01\x01\x12\x16\n\timage_uri\x18\x06 \x01(\tH\x05\x88\x01\x01\x12\x14\n\x0csecret_names\x18\x07 \x03(\t\x12<\n\x0fresource_limits\x18\x08 \x01(\x0b\x32\x1e.executor_api_pb.HostResourcesH\x06\x88\x01\x01\x42\x05\n\x03_idB\x0c\n\n_namespaceB\r\n\x0b_graph_nameB\x10\n\x0e_graph_versionB\x10\n\x0e_function_nameB\x0c\n\n_image_uriB\x12\n\x10_resource_limits"\xb8\x01\n\x15\x46unctionExecutorState\x12\x46\n\x0b\x64\x65scription\x18\x01 \x01(\x0b\x32,.executor_api_pb.FunctionExecutorDescriptionH\x00\x88\x01\x01\x12<\n\x06status\x18\x02 \x01(\x0e\x32\'.executor_api_pb.FunctionExecutorStatusH\x01\x88\x01\x01\x42\x0e\n\x0c_descriptionB\t\n\x07_status"\x9f\x05\n\rExecutorState\x12\x18\n\x0b\x65xecutor_id\x18\x01 \x01(\tH\x00\x88\x01\x01\x12\x1d\n\x10\x64\x65velopment_mode\x18\x02 \x01(\x08H\x01\x88\x01\x01\x12\x15\n\x08hostname\x18\x03 \x01(\tH\x02\x88\x01\x01\x12\x34\n\x06\x66lavor\x18\x04 \x01(\x0e\x32\x1f.executor_api_pb.ExecutorFlavorH\x03\x88\x01\x01\x12\x14\n\x07version\x18\x05 \x01(\tH\x04\x88\x01\x01\x12\x34\n\x06status\x18\x06 \x01(\x0e\x32\x1f.executor_api_pb.ExecutorStatusH\x05\x88\x01\x01\x12;\n\x0e\x66ree_resources\x18\x07 \x01(\x0b\x32\x1e.executor_api_pb.HostResourcesH\x06\x88\x01\x01\x12;\n\x11\x61llowed_functions\x18\x08 \x03(\x0b\x32 .executor_api_pb.AllowedFunction\x12H\n\x18\x66unction_executor_states\x18\t \x03(\x0b\x32&.executor_api_pb.FunctionExecutorState\x12:\n\x06labels\x18\n \x03(\x0b\x32*.executor_api_pb.ExecutorState.LabelsEntry\x12\x17\n\nstate_hash\x18\x0b \x01(\tH\x07\x88\x01\x01\x1a-\n\x0bLabelsEntry\x12\x0b\n\x03key\x18\x01 \x01(\t\x12\r\n\x05value\x18\x02 \x01(\t:\x02\x38\x01\x42\x0e\n\x0c_executor_idB\x13\n\x11_development_modeB\x0b\n\t_hostnameB\t\n\x07_flavorB\n\n\x08_versionB\t\n\x07_statusB\x11\n\x0f_free_resourcesB\r\n\x0b_state_hash"l\n\x1aReportExecutorStateRequest\x12;\n\x0e\x65xecutor_state\x18\x01 \x01(\x0b\x32\x1e.executor_api_pb.ExecutorStateH\x00\x88\x01\x01\x42\x11\n\x0f_executor_state"\x1d\n\x1bReportExecutorStateResponse"\x88\x03\n\x04Task\x12\x0f\n\x02id\x18\x01 \x01(\tH\x00\x88\x01\x01\x12\x16\n\tnamespace\x18\x02 \x01(\tH\x01\x88\x01\x01\x12\x17\n\ngraph_name\x18\x03 \x01(\tH\x02\x88\x01\x01\x12\x1a\n\rgraph_version\x18\x04 \x01(\tH\x03\x88\x01\x01\x12\x1a\n\rfunction_name\x18\x05 \x01(\tH\x04\x88\x01\x01\x12 \n\x13graph_invocation_id\x18\x06 \x01(\tH\x05\x88\x01\x01\x12\x16\n\tinput_key\x18\x08 \x01(\tH\x06\x88\x01\x01\x12\x1f\n\x12reducer_output_key\x18\t \x01(\tH\x07\x88\x01\x01\x12\x17\n\ntimeout_ms\x18\n \x01(\tH\x08\x88\x01\x01\x42\x05\n\x03_idB\x0c\n\n_namespaceB\r\n\x0b_graph_nameB\x10\n\x0e_graph_versionB\x10\n\x0e_function_nameB\x16\n\x14_graph_invocation_idB\x0c\n\n_input_keyB\x15\n\x13_reducer_output_keyB\r\n\x0b_timeout_ms"\x7f\n\x0eTaskAllocation\x12!\n\x14\x66unction_executor_id\x18\x01 \x01(\tH\x00\x88\x01\x01\x12(\n\x04task\x18\x02 \x01(\x0b\x32\x15.executor_api_pb.TaskH\x01\x88\x01\x01\x42\x17\n\x15_function_executor_idB\x07\n\x05_task"K\n\x1fGetDesiredExecutorStatesRequest\x12\x18\n\x0b\x65xecutor_id\x18\x01 \x01(\tH\x00\x88\x01\x01\x42\x0e\n\x0c_executor_id"\xb9\x01\n\x14\x44\x65siredExecutorState\x12H\n\x12\x66unction_executors\x18\x01 \x03(\x0b\x32,.executor_api_pb.FunctionExecutorDescription\x12\x39\n\x10task_allocations\x18\x02 \x03(\x0b\x32\x1f.executor_api_pb.TaskAllocation\x12\x12\n\x05\x63lock\x18\x03 \x01(\x04H\x00\x88\x01\x01\x42\x08\n\x06_clock*\x86\x03\n\x08GPUModel\x12\x15\n\x11GPU_MODEL_UNKNOWN\x10\x00\x12"\n\x1eGPU_MODEL_NVIDIA_TESLA_T4_16GB\x10\n\x12$\n GPU_MODEL_NVIDIA_TESLA_V100_16GB\x10\x14\x12\x1d\n\x19GPU_MODEL_NVIDIA_A10_24GB\x10\x1e\x12\x1f\n\x1bGPU_MODEL_NVIDIA_A6000_48GB\x10(\x12#\n\x1fGPU_MODEL_NVIDIA_A100_SXM4_40GB\x10\x32\x12#\n\x1fGPU_MODEL_NVIDIA_A100_SXM4_80GB\x10\x33\x12"\n\x1eGPU_MODEL_NVIDIA_A100_PCI_40GB\x10\x34\x12#\n\x1fGPU_MODEL_NVIDIA_H100_SXM5_80GB\x10<\x12"\n\x1eGPU_MODEL_NVIDIA_H100_PCI_80GB\x10=\x12"\n\x1eGPU_MODEL_NVIDIA_RTX_6000_24GB\x10>*\xa3\x03\n\x16\x46unctionExecutorStatus\x12$\n FUNCTION_EXECUTOR_STATUS_UNKNOWN\x10\x00\x12(\n$FUNCTION_EXECUTOR_STATUS_STARTING_UP\x10\x01\x12:\n6FUNCTION_EXECUTOR_STATUS_STARTUP_FAILED_CUSTOMER_ERROR\x10\x02\x12:\n6FUNCTION_EXECUTOR_STATUS_STARTUP_FAILED_PLATFORM_ERROR\x10\x03\x12!\n\x1d\x46UNCTION_EXECUTOR_STATUS_IDLE\x10\x04\x12)\n%FUNCTION_EXECUTOR_STATUS_RUNNING_TASK\x10\x05\x12&\n"FUNCTION_EXECUTOR_STATUS_UNHEALTHY\x10\x06\x12%\n!FUNCTION_EXECUTOR_STATUS_STOPPING\x10\x07\x12$\n FUNCTION_EXECUTOR_STATUS_STOPPED\x10\x08*\xc3\x01\n\x0e\x45xecutorStatus\x12\x1b\n\x17\x45XECUTOR_STATUS_UNKNOWN\x10\x00\x12\x1f\n\x1b\x45XECUTOR_STATUS_STARTING_UP\x10\x01\x12\x1b\n\x17\x45XECUTOR_STATUS_RUNNING\x10\x02\x12\x1b\n\x17\x45XECUTOR_STATUS_DRAINED\x10\x03\x12\x1c\n\x18\x45XECUTOR_STATUS_STOPPING\x10\x04\x12\x1b\n\x17\x45XECUTOR_STATUS_STOPPED\x10\x05*d\n\x0e\x45xecutorFlavor\x12\x1b\n\x17\x45XECUTOR_FLAVOR_UNKNOWN\x10\x00\x12\x17\n\x13\x45XECUTOR_FLAVOR_OSS\x10\x01\x12\x1c\n\x18\x45XECUTOR_FLAVOR_PLATFORM\x10\x02\x32\xff\x01\n\x0b\x45xecutorAPI\x12t\n\x15report_executor_state\x12+.executor_api_pb.ReportExecutorStateRequest\x1a,.executor_api_pb.ReportExecutorStateResponse"\x00\x12z\n\x1bget_desired_executor_states\x12\x30.executor_api_pb.GetDesiredExecutorStatesRequest\x1a%.executor_api_pb.DesiredExecutorState"\x00\x30\x01\x62\x06proto3'
+)
+
+_globals = globals()
+_builder.BuildMessageAndEnumDescriptors(DESCRIPTOR, _globals)
+_builder.BuildTopDescriptorsAndMessages(
+    DESCRIPTOR, "indexify.proto.executor_api_pb2", _globals
+)
+if not _descriptor._USE_C_DESCRIPTORS:
+    DESCRIPTOR._loaded_options = None
+    _globals["_EXECUTORSTATE_LABELSENTRY"]._loaded_options = None
+    _globals["_EXECUTORSTATE_LABELSENTRY"]._serialized_options = b"8\001"
+    _globals["_GPUMODEL"]._serialized_start = 2704
+    _globals["_GPUMODEL"]._serialized_end = 3094
+    _globals["_FUNCTIONEXECUTORSTATUS"]._serialized_start = 3097
+    _globals["_FUNCTIONEXECUTORSTATUS"]._serialized_end = 3516
+    _globals["_EXECUTORSTATUS"]._serialized_start = 3519
+    _globals["_EXECUTORSTATUS"]._serialized_end = 3714
+    _globals["_EXECUTORFLAVOR"]._serialized_start = 3716
+    _globals["_EXECUTORFLAVOR"]._serialized_end = 3816
+    _globals["_GPURESOURCES"]._serialized_start = 54
+    _globals["_GPURESOURCES"]._serialized_end = 155
+    _globals["_HOSTRESOURCES"]._serialized_start = 158
+    _globals["_HOSTRESOURCES"]._serialized_end = 352
+    _globals["_ALLOWEDFUNCTION"]._serialized_start = 355
+    _globals["_ALLOWEDFUNCTION"]._serialized_end = 542
+    _globals["_FUNCTIONEXECUTORDESCRIPTION"]._serialized_start = 545
+    _globals["_FUNCTIONEXECUTORDESCRIPTION"]._serialized_end = 910
+    _globals["_FUNCTIONEXECUTORSTATE"]._serialized_start = 913
+    _globals["_FUNCTIONEXECUTORSTATE"]._serialized_end = 1097
+    _globals["_EXECUTORSTATE"]._serialized_start = 1100
+    _globals["_EXECUTORSTATE"]._serialized_end = 1771
+    _globals["_EXECUTORSTATE_LABELSENTRY"]._serialized_start = 1608
+    _globals["_EXECUTORSTATE_LABELSENTRY"]._serialized_end = 1653
+    _globals["_REPORTEXECUTORSTATEREQUEST"]._serialized_start = 1773
+    _globals["_REPORTEXECUTORSTATEREQUEST"]._serialized_end = 1881
+    _globals["_REPORTEXECUTORSTATERESPONSE"]._serialized_start = 1883
+    _globals["_REPORTEXECUTORSTATERESPONSE"]._serialized_end = 1912
+    _globals["_TASK"]._serialized_start = 1915
+    _globals["_TASK"]._serialized_end = 2307
+    _globals["_TASKALLOCATION"]._serialized_start = 2309
+    _globals["_TASKALLOCATION"]._serialized_end = 2436
+    _globals["_GETDESIREDEXECUTORSTATESREQUEST"]._serialized_start = 2438
+    _globals["_GETDESIREDEXECUTORSTATESREQUEST"]._serialized_end = 2513
+    _globals["_DESIREDEXECUTORSTATE"]._serialized_start = 2516
+    _globals["_DESIREDEXECUTORSTATE"]._serialized_end = 2701
+    _globals["_EXECUTORAPI"]._serialized_start = 3819
+    _globals["_EXECUTORAPI"]._serialized_end = 4074
+# @@protoc_insertion_point(module_scope)

indexify/proto/{task_scheduler_pb2.pyi → executor_api_pb2.pyi}

@@ -50,6 +50,12 @@ class ExecutorStatus(int, metaclass=_enum_type_wrapper.EnumTypeWrapper):
     EXECUTOR_STATUS_STOPPING: _ClassVar[ExecutorStatus]
     EXECUTOR_STATUS_STOPPED: _ClassVar[ExecutorStatus]
 
+class ExecutorFlavor(int, metaclass=_enum_type_wrapper.EnumTypeWrapper):
+    __slots__ = ()
+    EXECUTOR_FLAVOR_UNKNOWN: _ClassVar[ExecutorFlavor]
+    EXECUTOR_FLAVOR_OSS: _ClassVar[ExecutorFlavor]
+    EXECUTOR_FLAVOR_PLATFORM: _ClassVar[ExecutorFlavor]
+
 GPU_MODEL_UNKNOWN: GPUModel
 GPU_MODEL_NVIDIA_TESLA_T4_16GB: GPUModel
 GPU_MODEL_NVIDIA_TESLA_V100_16GB: GPUModel
@@ -76,6 +82,9 @@ EXECUTOR_STATUS_RUNNING: ExecutorStatus
 EXECUTOR_STATUS_DRAINED: ExecutorStatus
 EXECUTOR_STATUS_STOPPING: ExecutorStatus
 EXECUTOR_STATUS_STOPPED: ExecutorStatus
+EXECUTOR_FLAVOR_UNKNOWN: ExecutorFlavor
+EXECUTOR_FLAVOR_OSS: ExecutorFlavor
+EXECUTOR_FLAVOR_PLATFORM: ExecutorFlavor
 
 class GPUResources(_message.Message):
     __slots__ = ("count", "model")
@@ -178,30 +187,59 @@ class ExecutorState(_message.Message):
     __slots__ = (
         "executor_id",
         "development_mode",
-        "executor_status",
+        "hostname",
+        "flavor",
+        "version",
+        "status",
         "free_resources",
         "allowed_functions",
         "function_executor_states",
+        "labels",
+        "state_hash",
     )
+
+    class LabelsEntry(_message.Message):
+        __slots__ = ("key", "value")
+        KEY_FIELD_NUMBER: _ClassVar[int]
+        VALUE_FIELD_NUMBER: _ClassVar[int]
+        key: str
+        value: str
+        def __init__(
+            self, key: _Optional[str] = ..., value: _Optional[str] = ...
+        ) -> None: ...
+
     EXECUTOR_ID_FIELD_NUMBER: _ClassVar[int]
     DEVELOPMENT_MODE_FIELD_NUMBER: _ClassVar[int]
-    EXECUTOR_STATUS_FIELD_NUMBER: _ClassVar[int]
+    HOSTNAME_FIELD_NUMBER: _ClassVar[int]
+    FLAVOR_FIELD_NUMBER: _ClassVar[int]
+    VERSION_FIELD_NUMBER: _ClassVar[int]
+    STATUS_FIELD_NUMBER: _ClassVar[int]
     FREE_RESOURCES_FIELD_NUMBER: _ClassVar[int]
     ALLOWED_FUNCTIONS_FIELD_NUMBER: _ClassVar[int]
     FUNCTION_EXECUTOR_STATES_FIELD_NUMBER: _ClassVar[int]
+    LABELS_FIELD_NUMBER: _ClassVar[int]
+    STATE_HASH_FIELD_NUMBER: _ClassVar[int]
     executor_id: str
     development_mode: bool
-    executor_status: ExecutorStatus
+    hostname: str
+    flavor: ExecutorFlavor
+    version: str
+    status: ExecutorStatus
     free_resources: HostResources
     allowed_functions: _containers.RepeatedCompositeFieldContainer[AllowedFunction]
     function_executor_states: _containers.RepeatedCompositeFieldContainer[
         FunctionExecutorState
     ]
+    labels: _containers.ScalarMap[str, str]
+    state_hash: str
     def __init__(
         self,
         executor_id: _Optional[str] = ...,
         development_mode: bool = ...,
-        executor_status: _Optional[_Union[ExecutorStatus, str]] = ...,
+        hostname: _Optional[str] = ...,
+        flavor: _Optional[_Union[ExecutorFlavor, str]] = ...,
+        version: _Optional[str] = ...,
+        status: _Optional[_Union[ExecutorStatus, str]] = ...,
         free_resources: _Optional[_Union[HostResources, _Mapping]] = ...,
         allowed_functions: _Optional[
             _Iterable[_Union[AllowedFunction, _Mapping]]
@@ -209,6 +247,8 @@ class ExecutorState(_message.Message):
         function_executor_states: _Optional[
             _Iterable[_Union[FunctionExecutorState, _Mapping]]
         ] = ...,
+        labels: _Optional[_Mapping[str, str]] = ...,
+        state_hash: _Optional[str] = ...,
     ) -> None: ...
 
 class ReportExecutorStateRequest(_message.Message):

indexify/proto/{task_scheduler_pb2_grpc.py → executor_api_pb2_grpc.py}

@@ -4,9 +4,7 @@ import warnings
 
 import grpc
 
-from indexify.proto import (
-    task_scheduler_pb2 as indexify_dot_proto_dot_task__scheduler__pb2,
-)
+from indexify.proto import executor_api_pb2 as indexify_dot_proto_dot_executor__api__pb2
 
 GRPC_GENERATED_VERSION = "1.70.0"
 GRPC_VERSION = grpc.__version__
@@ -24,16 +22,20 @@ except ImportError:
 if _version_not_supported:
     raise RuntimeError(
         f"The grpc package installed is at version {GRPC_VERSION},"
-        + f" but the generated code in indexify/proto/task_scheduler_pb2_grpc.py depends on"
+        + f" but the generated code in indexify/proto/executor_api_pb2_grpc.py depends on"
         + f" grpcio>={GRPC_GENERATED_VERSION}."
         + f" Please upgrade your grpc module to grpcio>={GRPC_GENERATED_VERSION}"
         + f" or downgrade your generated code using grpcio-tools<={GRPC_VERSION}."
     )
 
 
-class TaskSchedulerServiceStub(object):
+class ExecutorAPIStub(object):
     """Internal API for scheduling and running tasks on Executors. Executors are acting as clients of this API.
     Server is responsible for scheduling tasks on Executors and Executors are responsible for running the tasks.
+
+    Rename with caution. Existing clients won't find the service if the service name changes. A HTTP2 ingress proxy
+    might use the service name in it HTTP2 path based routing rules. See how gRPC uses service names in its HTTP2 paths
+    at https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md.
     """
 
     def __init__(self, channel):
@@ -43,22 +45,26 @@ class TaskSchedulerServiceStub(object):
             channel: A grpc.Channel.
         """
         self.report_executor_state = channel.unary_unary(
-            "/task_scheduler_service.TaskSchedulerService/report_executor_state",
-            request_serializer=indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateRequest.SerializeToString,
-            response_deserializer=indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateResponse.FromString,
+            "/executor_api_pb.ExecutorAPI/report_executor_state",
+            request_serializer=indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateRequest.SerializeToString,
+            response_deserializer=indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateResponse.FromString,
             _registered_method=True,
         )
         self.get_desired_executor_states = channel.unary_stream(
-            "/task_scheduler_service.TaskSchedulerService/get_desired_executor_states",
-            request_serializer=indexify_dot_proto_dot_task__scheduler__pb2.GetDesiredExecutorStatesRequest.SerializeToString,
-            response_deserializer=indexify_dot_proto_dot_task__scheduler__pb2.DesiredExecutorState.FromString,
+            "/executor_api_pb.ExecutorAPI/get_desired_executor_states",
+            request_serializer=indexify_dot_proto_dot_executor__api__pb2.GetDesiredExecutorStatesRequest.SerializeToString,
+            response_deserializer=indexify_dot_proto_dot_executor__api__pb2.DesiredExecutorState.FromString,
             _registered_method=True,
         )
 
 
-class TaskSchedulerServiceServicer(object):
+class ExecutorAPIServicer(object):
     """Internal API for scheduling and running tasks on Executors. Executors are acting as clients of this API.
     Server is responsible for scheduling tasks on Executors and Executors are responsible for running the tasks.
+
+    Rename with caution. Existing clients won't find the service if the service name changes. A HTTP2 ingress proxy
+    might use the service name in it HTTP2 path based routing rules. See how gRPC uses service names in its HTTP2 paths
+    at https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md.
     """
 
     def report_executor_state(self, request, context):
@@ -81,32 +87,36 @@ class TaskSchedulerServiceServicer(object):
         raise NotImplementedError("Method not implemented!")
 
 
-def add_TaskSchedulerServiceServicer_to_server(servicer, server):
+def add_ExecutorAPIServicer_to_server(servicer, server):
     rpc_method_handlers = {
         "report_executor_state": grpc.unary_unary_rpc_method_handler(
             servicer.report_executor_state,
-            request_deserializer=indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateRequest.FromString,
-            response_serializer=indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateResponse.SerializeToString,
+            request_deserializer=indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateRequest.FromString,
+            response_serializer=indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateResponse.SerializeToString,
         ),
         "get_desired_executor_states": grpc.unary_stream_rpc_method_handler(
             servicer.get_desired_executor_states,
-            request_deserializer=indexify_dot_proto_dot_task__scheduler__pb2.GetDesiredExecutorStatesRequest.FromString,
-            response_serializer=indexify_dot_proto_dot_task__scheduler__pb2.DesiredExecutorState.SerializeToString,
+            request_deserializer=indexify_dot_proto_dot_executor__api__pb2.GetDesiredExecutorStatesRequest.FromString,
+            response_serializer=indexify_dot_proto_dot_executor__api__pb2.DesiredExecutorState.SerializeToString,
         ),
     }
     generic_handler = grpc.method_handlers_generic_handler(
-        "task_scheduler_service.TaskSchedulerService", rpc_method_handlers
+        "executor_api_pb.ExecutorAPI", rpc_method_handlers
     )
     server.add_generic_rpc_handlers((generic_handler,))
     server.add_registered_method_handlers(
-        "task_scheduler_service.TaskSchedulerService", rpc_method_handlers
+        "executor_api_pb.ExecutorAPI", rpc_method_handlers
     )
 
 
 # This class is part of an EXPERIMENTAL API.
-class TaskSchedulerService(object):
+class ExecutorAPI(object):
     """Internal API for scheduling and running tasks on Executors. Executors are acting as clients of this API.
     Server is responsible for scheduling tasks on Executors and Executors are responsible for running the tasks.
+
+    Rename with caution. Existing clients won't find the service if the service name changes. A HTTP2 ingress proxy
+    might use the service name in it HTTP2 path based routing rules. See how gRPC uses service names in its HTTP2 paths
+    at https://github.com/grpc/grpc/blob/master/doc/PROTOCOL-HTTP2.md.
     """
 
     @staticmethod
@@ -125,9 +135,9 @@ class TaskSchedulerService(object):
         return grpc.experimental.unary_unary(
             request,
             target,
-            "/task_scheduler_service.TaskSchedulerService/report_executor_state",
-            indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateRequest.SerializeToString,
-            indexify_dot_proto_dot_task__scheduler__pb2.ReportExecutorStateResponse.FromString,
+            "/executor_api_pb.ExecutorAPI/report_executor_state",
+            indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateRequest.SerializeToString,
+            indexify_dot_proto_dot_executor__api__pb2.ReportExecutorStateResponse.FromString,
             options,
             channel_credentials,
             insecure,
@@ -155,9 +165,9 @@ class TaskSchedulerService(object):
         return grpc.experimental.unary_stream(
             request,
             target,
-            "/task_scheduler_service.TaskSchedulerService/get_desired_executor_states",
-            indexify_dot_proto_dot_task__scheduler__pb2.GetDesiredExecutorStatesRequest.SerializeToString,
-            indexify_dot_proto_dot_task__scheduler__pb2.DesiredExecutorState.FromString,
+            "/executor_api_pb.ExecutorAPI/get_desired_executor_states",
+            indexify_dot_proto_dot_executor__api__pb2.GetDesiredExecutorStatesRequest.SerializeToString,
+            indexify_dot_proto_dot_executor__api__pb2.DesiredExecutorState.FromString,
             options,
             channel_credentials,
             insecure,

{indexify-0.3.16.dist-info → indexify-0.3.18.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: indexify
-Version: 0.3.16
+Version: 0.3.18
 Summary: Open Source Indexify components and helper tools
 Home-page: https://github.com/tensorlakeai/indexify
 License: Apache 2.0