Skip to content

Commit

Permalink
Add docs for building with debug symbols (triton-inference-server#3153)
Browse files Browse the repository at this point in the history
* Add docs for debug build

* Add segfault doc to faq

* Update copyright

* Add comment for including bt in issue

* Review edits
  • Loading branch information
CoderHam authored Jul 27, 2021
1 parent a3faff7 commit d70a8a8
Show file tree
Hide file tree
Showing 2 changed files with 19 additions and 2 deletions.
8 changes: 7 additions & 1 deletion docs/build.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
<!--
# Copyright (c) 2018-2021, NVIDIA CORPORATION. All rights reserved.
# Copyright 2018-2021, 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
Expand Down Expand Up @@ -336,3 +336,9 @@ changes in the following areas.
NGC container. But the build can also use PyTorch shared libraries
that you build separately for your platform. See the pytorch_backend
build process for details.

## Building with Debug Symbols

To build with Debug symbols, use the --build-type=Debug arguement while
launching build.py. You can then launch the built server with gdb and see
the debug symbols/information in the gdb trace.
13 changes: 12 additions & 1 deletion docs/faq.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
<!--
# Copyright (c) 2019-2020, NVIDIA CORPORATION. All rights reserved.
# Copyright 2019-2021, 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
Expand Down Expand Up @@ -149,3 +149,14 @@ server into multiple *nodes*, each with one GPU. In this case the
orchestration environment will run a different Triton for each GPU and
an load balancer will be used to divide inference requests across the
available Triton instances.

## If the server segfaults, how can I debug it?

The NGC build is a Release build and does not contain Debug symbols.
The build.py as well defaults to a Release build. Refer to the instructions
in [build.md](build.md#building-with-debug-symbols) to create a Debug build
of Triton. This will help find the cause of the segmentation fault when
looking at the gdb trace for the segfault.

When opening a GitHub issue for the segfault with Triton, please include
the backtrace to better help us resolve the problem.

0 comments on commit d70a8a8

Please sign in to comment.