HTTP live connections on server shutdown (#335)

* Add TRITONSERVER_ServerStopWithTimeout API

* Use TRITONSERVER_ServerSetExitTimeout() API

* TRITONSERVER_API_VERSION

* Enhance c api descriptions

Co-authored-by: Neelay Shah <[email protected]>

* Fix pre-commit for 'Enhance c api descriptions'

---------

Co-authored-by: Neelay Shah <[email protected]>

Author: kthui

include/triton/core/tritonserver.h

@@ -91,7 +91,7 @@ struct TRITONSERVER_MetricFamily;
 ///   }
 ///
 #define TRITONSERVER_API_VERSION_MAJOR 1
-#define TRITONSERVER_API_VERSION_MINOR 29
+#define TRITONSERVER_API_VERSION_MINOR 30
 
 /// Get the TRITONBACKEND API version supported by the Triton shared
 /// library. This value can be compared against the
@@ -2258,6 +2258,17 @@ TRITONSERVER_DECLSPEC struct TRITONSERVER_Error* TRITONSERVER_ServerDelete(
 TRITONSERVER_DECLSPEC struct TRITONSERVER_Error* TRITONSERVER_ServerStop(
     struct TRITONSERVER_Server* server);
 
+/// Set the exit timeout on the server object. This value overrides the value
+/// initially set through server options and provides a mechanism to update the
+/// exit timeout while the serving is running.
+///
+/// \param server The inference server object.
+/// \param timeout The exit timeout, in seconds.
+/// \return a TRITONSERVER_Error indicating success or failure.
+TRITONSERVER_DECLSPEC struct TRITONSERVER_Error*
+TRITONSERVER_ServerSetExitTimeout(
+    struct TRITONSERVER_Server* server, unsigned int timeout);
+
 /// Register a new model repository. Not available in polling mode.
 ///
 /// \param server The inference server object.

src/tritonserver.cc

@@ -2572,6 +2572,17 @@ TRITONSERVER_ServerStop(TRITONSERVER_Server* server)
   return nullptr;  // Success
 }
 
+TRITONAPI_DECLSPEC TRITONSERVER_Error*
+TRITONSERVER_ServerSetExitTimeout(
+    TRITONSERVER_Server* server, unsigned int timeout)
+{
+  tc::InferenceServer* lserver = reinterpret_cast<tc::InferenceServer*>(server);
+  if (lserver != nullptr) {
+    lserver->SetExitTimeoutSeconds(timeout);
+  }
+  return nullptr;  // Success
+}
+
 TRITONSERVER_DECLSPEC TRITONSERVER_Error*
 TRITONSERVER_ServerRegisterModelRepository(
     TRITONSERVER_Server* server, const char* repository_path,

src/tritonserver_stub.cc

@@ -559,6 +559,10 @@ TRITONSERVER_ServerStop()
 {
 }
 TRITONAPI_DECLSPEC void
+TRITONSERVER_ServerSetExitTimeout()
+{
+}
+TRITONAPI_DECLSPEC void
 TRITONSERVER_ServerPollModelRepository()
 {
 }