Fix bugs in GRPC handler changes (#3823)

triton-inference-server · Jan 19, 2022 · 4f925a7 · 4f925a7
1 parent f8f6d7b
commit 4f925a7
Show file tree

Hide file tree

Showing 3 changed files with 24 additions and 21 deletions.
diff --git a/qa/L0_cmdline_trace/test.sh b/qa/L0_cmdline_trace/test.sh
@@ -1,5 +1,5 @@
 #!/bin/bash
-# Copyright (c) 2019-2021, NVIDIA CORPORATION. All rights reserved.
+# Copyright (c) 2019-2022, NVIDIA CORPORATION. All rights reserved.
 #
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions
@@ -302,16 +302,18 @@ if [ `grep -c "COMPUTE_INPUT_END" summary_ensemble.log` != "7" ]; then
     echo -e "Ensemble trace log expects 7 compute"
     RET=1
 fi
+# For GRPC frontend, its handlers will occupy one trace ID on creation
+GRPC_ID_OFFSET=3
 for trace_str in \
-        "{\"id\":3,\"model_name\":\"simple\",\"model_version\":1}" \
-        "{\"id\":4,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":3}" \
-        "{\"id\":5,\"model_name\":\"fan_${MODELBASE}\",\"model_version\":1,\"parent_id\":3}" \
-        "{\"id\":6,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":5}" \
-        "{\"id\":7,\"model_name\":\"${MODELBASE}\",\"model_version\":1,\"parent_id\":5}" \
-        "{\"id\":8,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":5}" \
-        "{\"id\":9,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":5}" \
-        "{\"id\":10,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":3}" \
-        "{\"id\":11,\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":3}" ; do
+        "{\"id\":$((GRPC_ID_OFFSET+1)),\"model_name\":\"simple\",\"model_version\":1}" \
+        "{\"id\":$((GRPC_ID_OFFSET+2)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+1))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+3)),\"model_name\":\"fan_${MODELBASE}\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+1))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+4)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+3))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+5)),\"model_name\":\"${MODELBASE}\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+3))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+6)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+3))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+7)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+3))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+8)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+1))}" \
+        "{\"id\":$((GRPC_ID_OFFSET+9)),\"model_name\":\"nop_TYPE_INT32_-1\",\"model_version\":1,\"parent_id\":$((GRPC_ID_OFFSET+1))}" ; do
     if [ `grep -c ${trace_str} trace_ensemble.log` != "1" ]; then
         echo -e "Ensemble trace log expects trace: ${trace_str}"
         RET=1

diff --git a/src/servers/grpc_server.cc b/src/servers/grpc_server.cc
@@ -1,4 +1,4 @@
-// Copyright 2019-2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// Copyright 2019-2022, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 //
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions
@@ -4179,17 +4179,18 @@ GRPCServer::Start()
         compression_level_);
     hmodelinfer->Start();
     model_infer_handlers_.emplace_back(hmodelinfer);
-
-    // Handler for streaming inference requests.
-    ModelStreamInferHandler* hmodelstreaminfer = new ModelStreamInferHandler(
-        "ModelStreamInferHandler", server_, trace_manager_, shm_manager_,
-        &service_, model_stream_infer_cq_.get(),
-        infer_allocation_pool_size_ /* max_state_bucket_count */,
-        compression_level_);
-    hmodelstreaminfer->Start();
-    model_stream_infer_handlers_.emplace_back(hmodelstreaminfer);
   }
 
+  // Handler for streaming inference requests. Keeps one handler for streaming
+  // to avoid possible concurrent writes which is not allowed
+  ModelStreamInferHandler* hmodelstreaminfer = new ModelStreamInferHandler(
+      "ModelStreamInferHandler", server_, trace_manager_, shm_manager_,
+      &service_, model_stream_infer_cq_.get(),
+      infer_allocation_pool_size_ /* max_state_bucket_count */,
+      compression_level_);
+  hmodelstreaminfer->Start();
+  model_stream_infer_handlers_.emplace_back(hmodelstreaminfer);
+
   running_ = true;
   LOG_INFO << "Started GRPCInferenceService at " << server_addr_;
   return nullptr;  // success

diff --git a/src/servers/grpc_server.h b/src/servers/grpc_server.h
@@ -1,4 +1,4 @@
-// Copyright 2019-2021, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+// Copyright 2019-2022, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 //
 // Redistribution and use in source and binary forms, with or without
 // modification, are permitted provided that the following conditions