# root @ snode0 in /etc/systemd/system [17:57:48] $ journalctl -u webhook.service -- Logs begin at Mon 2022-06-06 15:54:50 CST, end at Tue 2022-06-28 17:57:50 CST. -- Jun 28 17:30:53 snode0 systemd[1]: Started Webhook receiver for GitHub.
问题
1 2 3 4 5 6 7 8 9 10
$ systemctl reload webhook.service ==== AUTHENTICATING FOR org.freedesktop.systemd1.manage-units === Authentication is required to reload 'webhook.service'. Multiple identities can be used for authentication: 1. Jun Shi (shijun) 2. Shaojie Tan (shaojiemike) Choose identity to authenticate as (1-2): 2 Password: ==== AUTHENTICATION COMPLETE === Failed to reload webhook.service: Job type reload is not applicable for unit webhook.service.
All developers should start with Nsight Systems to identify the largest optimization opportunities. Nsight Systems provides developers a system-wide visualization of an applications performance. Developers can optimize bottlenecks to scale efficiently across any number or size of CPUs and GPUs; from large servers to our smallest SoC. For further optimizations to compute kernels developers should use Nsight Compute or to further optimize a graphics workloads, use Nsight Graphics.
Nsight Compute
Nsight Compute is an interactive kernel profiler for CUDA applications. It provides detailed performance metrics and API debugging via a user interface and command line tool. Nsight Compute also provides customizable and data-driven user interface and metric collection that can be extended with analysis scripts for post-processing results.
Nsight Graphics
Nsight Graphics is a standalone application for the debugging, profiling, and analysis of graphics applications on Microsoft Windows and Linux. It allows you to optimize the performance of your Direct3D 11, Direct3D 12, DirectX Raytracing 1.1, OpenGL, Vulkan, and KHR Vulkan Ray Tracing Extension based applications.
Install Nsight local
check the perf config To collect thread scheduling data and IP (instruction pointer) samples
cat /proc/sys/kernel/perf_event_paranoid
如果大于2，临时改变 sudo sh -c 'echo 2 >/proc/sys/kernel/perf_event_paranoid'重启会重置
永久修改 sudo sh -c 'echo kernel.perf_event_paranoid=2 > /etc/sysctl.d/local.conf'
GPU Metrics [0]: The user running Nsight Systems does not have permission to access NVIDIA GPU Performance Counters on the target device. For more details, please visit https://developer.nvidia.com/ERR_NVGPUCTRPERM - API function: NVPW_GPU_PeriodicSampler_GetCounterAvailability(¶ms) - Error code: 17 - Source function: static std::vector<unsigned char> QuadDDaemon::EventSource::GpuMetricsBackend::Impl::CounterConfig::GetCounterAvailabilityImage(uint32_t) - Source location: /dvs/p4/build/sw/devtools/Agora/Rel/DTC_F/QuadD/Target/quadd_d/quadd_d/jni/EventSource/GpuMetricsBackend.cpp:609
Profile 速度
大致2到3倍时间：默认采样率，单独运行52s, Nsight-sys模拟需要135s。
HPC APP : PCIE, GPU DRAM Bandwidth, Warp
GPU Metrics选项能看出 PCIE, GPU DRAM Bandwidth, Warp的使用情况。
Compute Warps in Flight
将鼠标放在上面会有具体的数值或者名称的解释，(正在使用的Warps)
Unallocated Warps in Active SMs
Definition: This metric represents the number of warps that are not actively executing but are assigned to an active Streaming Multiprocessor (SM).
Interpretation: In CUDA, SMs are the fundamental processing units on the GPU. Each SM can execute multiple warps concurrently. “Unallocated Warps in Active SMs” indicates the number of warps that are ready to be scheduled on an SM but are currently waiting due to resource contention or other factors. A high number may suggest that there is room for additional work but available resources are not fully utilized.
这个错误也是令人迷惑 The memory access pattern for loads from L1TEX to L2 is not optimal. The granularity of an L1TEX request to L2 is a 128 byte cache line. That is 4 consecutive 32-byte sectors per L2 request. However, this kernel only accesses an average of 3.7 sectors out of the possible 4 sectors per cache line. Check the Source Counters section for uncoalesced loads and try to minimize how many cache lines need to be accessed per memory request.
Warning: Failed to get OpenGL version. OpenGL version 2.0 or higher is required. OpenGL version is too low (0). Falling back to Mesa software rendering. qt.qpa.plugin: Could not load the Qt platform plugin "xcb"in"" even though it was found. This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.
Available platform plugins are: offscreen, wayland-egl, wayland, wayland-xcomposite-egl, wayland-xcomposite-glx, xcb.
qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found. This application failed to start because no Qt platform plugin could be initialized. Reinstalling the application may fix this problem.
Available platform plugins are: xcb.
Application could not be initialized! This is likely due to missing Qt platform dependencies. For a list of dependencies, please refer to https://doc.qt.io/qt-5/linux-requirements.html To view missing libraries, set QT_DEBUG_PLUGINS=1 and re-run the application.
按照说明 export QT_DEBUG_PLUGINS=1再次运行, 显示具体问题
1
Cannot load library /staff/shaojiemike/Install/cuda_11.7.0_515.43.04_linux/nsight-compute-2022.2.0/host/linux-desktop-glibc_2_11_3-x64/Plugins/platforms/libqxcb.so: (libxcb-xinput.so.0: cannot open shared object file: No such file or directory)
解决 sudo apt-get install libxcb-xinput0
kernel没权限profile
ERR_NVGPUCTRPERM - The user does not have permission to profile on the target device
要用sudo，或者最新的NV
could not connect to display localhost:10.0 under sudo
1 2 3 4 5 6
$ sudo ncu-ui MobaXterm X11 proxy: Authorisation not recognised qt.qpa.xcb: could not connect to display localhost:10.0
how docker run different ubuntu version Sharing the same running kernel? no SW conficts?
If your host kernel is “compatible enough“ with the software in the container you want to run it will work; otherwise, it won’t.[^1] So what does “compatible enough” mean? It depends on what requests the program makes of the kernel (system calls) and** what features it expects the kernel to support**. Some programs make requests that will break things; others don’t.
Compatibility
on an Ubuntu 18.04 (kernel 4.19) or similar host:[^1] docker run centos:7 bash works fine. docker run centos:6 bash fails with exit code 139, meaning it terminated with a segmentation violation signal; this is because the 4.19 kernel doesn’t support something that that build of bash tried to do. docker run centos:6 ls works fine because it’s not making a request the kernel can’t handle, as bash was. If you try docker run centos:6 bash on an older kernel, say 4.9 or earlier, you’ll find it will work fine.
Portainer 统一管理
安装Portainer Community Edition (CE)而不是Portainer Business Edition (BE)
Image like a static configed/compiled software using dockerfile/gcc.
And container is a running process that we can control.
Dockerfile
First, write an installation script for all of your dependencies. This script is written with Docker specific syntax and is called a Dockerfile(1). { .annotate }
A Dockerfile is a script used to create a Docker image, which is a lightweight, standalone, and executable package that includes everything needed to run a piece of software, including the code, runtime, system tools, system libraries, and settings.
Building image using Dockerfile
Using dockerfile in portainer
`Images > Build image`
Dockerfile
Here’s a simple Dockerfile code snippet(it’s usually built on official base image):
# Set environment variables ENV MY_ENV_VARIABLE my_value
# Run commands to install packages and set up the environment RUN apt-get update && apt-get install -y \ package1 \ package2 \ && rm -rf /var/lib/apt/lists/*
# Copy files from your local machine to the container COPY local_directory /container_directory
# Set the working directory WORKDIR /app
# Expose a port EXPOSE8080
# Define the command to run when the container starts CMD ["command_to_start_application"]
Codeforces: Another good platform with nice contests where you will get to learn new things that will improve your understanding of concepts. It’s ratings don’t matter to recruiters but the things you’ll learn will help you during interviews and company’s coding tests.
TOPCODER: This caters to every aspect of software develpment. They have both designing and development contest. Major companies provide Topcoder community to experment with their nw products or APIs. But if you want to mention it in your CV, your ratings should be above 1400 to impress a recruiter.
{ align=left }
{ align=right }
[^2] 
