Thursday, September 16, 2021

BPFTrace 範例

BPFTrace
Install and compile
sudo apt-get update
sudo apt-get install -y bison cmake flex g++ git libelf-dev zlib1g-dev libfl-dev systemtap-sdt-dev binutils-dev
sudo apt-get install -y llvm-7-dev llvm-7-runtime libclang-7-dev clang-7
git clone https://github.com/iovisor/bpftrace
mkdir bpftrace/build; cd bpftrace/build;
cmake -DCMAKE_BUILD_TYPE=Release ..
make -j8

make install 

TCP/UDP Tracing

sudo bpftrace -e "$(cat udp_sendmsg2.bt)"
sudo bpftrace -e "$(cat udpconnect.bt)"
sudo python trace.py -I 'net/sock.h' 'udp_sendmsg(struct sock *sk) "daddr: %s, dport: %d", sk->sk_daddr, sk->sk_dport'

sudo bpftrace -e 'k:tcp_sendmsg,k:tcp_recvmsg {@[func,pid,tid,comm,args] = count();}'
sudo bpftrace -e 'k:tcp_sendmsg,k:tcp_recvmsg { @[func, pid, tid, comm] = count(); }'
sudo tcpdump -i any 'dst host 10.217.1.80 or dst host 10.217.1.50 or dst host 10.103.221.119 or dst host 10.110.176.166'

Golang 設定與常用指令說明

 Golang設定

加入下列於 ~/.profile
#Golang path
export PATH=$PATH:/usr/local/go/bin
export GOPATH=$HOME/go
export PATH=$PATH:$GOPATH/bin


常用的指令與使用說明
初始化 Module
這個會在資料夾底下產生一個初始的 go.mod 檔案
go mod init <module>
  • 在 build go 專案的時候,會根據 module name 的最後一路路徑 myproject 當作輸出檔名,會產生 myproject.exe 的檔案
  • 在 import 自己專案 package 的路徑的時候,github.com/mileslin/myproject 是根目錄,例如我新增一個 pg1 的 package ,則我在 import 的時候要輸入 import "github.com/mileslin/myproject/pg1"。

取得套件
go get -u github.com/gorilla/mux (-u 是說明,取得套件的時候不要從 cache 取得

#go get 的時候指定版本
go get github.com/gorilla/mux@v1.6.1
除了會新增一個 go.sum 檔案之外呢, go.mod 也會記錄你下載的套件,如下範例
module github.com/mileslin/myproject
go 1.13
require github.com/gorilla/mux v1.7.3 // indirect
// indirect 的意思是說,此套件還沒被用到。 所以只要在程式碼中使用此到件,並且 build 專案後,// indirect 就會消失
最後提一下,下載的套件會被放在 GOPATH/pkg/mod 底下。 (可以使用 go env 指令查看 GOPATH 路徑)

列出套件與套件版本
go list 取得可以使用的套件列表
$ go list all 會列出所有套件,包含 GOROOT 套件
$ go list -m all 只列出 module 列出的套件(包含自己的 package)
$ go list -m -versions github.com/gorilla/mux 列出此套件的所有版本

檢查套件使用情形
列出使套件在哪裡有被用到
$ go mod why github.com/gorilla/mux

Module download
這個會將 go.mod 定義的 library 都下載下來,但基本上 go build, go test 也會自動下載 library 並新增到 go.mod 上,
並會產出 go.sum 來確保 library 是否有更新。可以說能達到一樣的效果
go mod download

如果你忘了 import ,則會出現以下的錯誤
// github.com/streadway/amqp
go build -o ./bin/rabbitmq ./cmd
# rabbitmq/cmd
cmd/main.go:16:15: undefined: amqp
cmd/main.go:41:4: undefined: amqp

Module tidy
移除在 go.mod 不需要的 library
go mod tidy


go test
go test -timeout 30s ./probe/endpoint/... -run "TestSpyNoProcesses|TestSpyWithProcesses"
go test -timeout 30s ./probe/endpoint/...
go test -v -run=TestNewWithAssert ./example18-write-testing-and-doc/...

inspektor gadget 常用commands

 編譯與安裝kubectl-gadget

git clone https://github.com/kinvolk/inspektor-gadget.git
cd inspektor-gadget
make kubectl-gadget-linux-amd64
sudo cp kubectl-gadget-linux-amd64 /usr/local/bin/kubectl-gadget

佈署Gadgets on K8S 使用kubectl gadget
kubectl gadget deploy | kubectl apply -f -

or
#指定image
kubectl gadget deploy --image=kinvolk/gadget:fix63 | kubectl apply -f -
# make build & get image kinvolk/gadget:alban-fix63
kubectl gadget deploy --image=kinvolk/gadget:alban-fix63 | kubectl apply -f -

查詢在每個Node上的 gadget's pod name
kubectl get pod -n kube-system -l k8s-app=gadget -o=jsonpath='{.items[*].metadata.name}'
kubectl exec -it <gadget-xxxx> -n kube-system -- bash

Run BCC commands
$ kubectl gadget tcpconnect -A
$ kubectl gadget tcptracer -A


Traceloop
查看traceloop 清單
kubectl gadget traceloop list
NODE NAMESPACE PODNAME PODUID INDEX TRACEID CONTAINERID STATUS
ubuntu1804-k8s1 kube-system hubble-bdqn9 851a4185 0 0000b7ec718351b6 b2d36529 created 6 days ago
ubuntu1804-k8s1 default deathstar-657477f57d-crf8w bac896ec 0 0000b7f1f2b1d065 941b6b07 created 6 days ago
ubuntu1804-k8s2 kube-system coredns-6955765f44-nqr56 024f3577 0 0000541b48333690 62d1fade created 6 days ago
ubuntu1804-k8s2 kube-system coredns-6955765f44-pbwt5 00be4172 0 0000541b483869c5 59c000af created 6 days ago
ubuntu1804-k8s2 kube-system hubble-ui-7b767c4654-j7fs7 50ab917c 0 0000541b77264f34 dc9d5823 created 6 days ago
ubuntu1804-k8s2 default helloworld-golang cdb85f1e 1 0000eea8600e2b9a 82dfe69f created 4 days ago
ubuntu1804-k8s2 default helloworld-python 45139ac8 0 0000eea7d9821155 9147da4f created 4 days ago
ubuntu1804-k8s2 default deathstar-657477f57d-pflwg 63bd1b5d 0 00005421e8b30d05 fa336b14 created 6 days ago
ubuntu1804-k8s2 default helloworld-golang cdb85f1e 0 0000eeb0142b9961 b07b2652 created 4 days ago

traceloop 相關 commands
kubectl gadget traceloop pod default helloworld-golang 1
kubectl gadget traceloop show 0004ac1f39174f40

kubectl gadget traceloop pod sock-shop front-end-85c54f4fb-wftx6 0

# 須進入到gadget pod內才能使用
curl --silent --unix-socket /run/traceloop.socket http://localhost/dump-by-traceid?traceid=0000b7ec718351b6


Network Policy
Blog 介紹Network Policy Advisor

執行 network-policy monitor command
kubectl gadget network-policy monitor --namespaces default --output ./networktrace.log

製造traffic by curl
#For instance:
kubectl exec helloworld-python -- curl -s -XGET http://helloworld-golang.default.svc.cluster.local:8001/
kubectl exec xwing -- curl -s -XGET http://helloworld-python.default.svc.cluster.local:8080/

產生network-policy
kubectl gadget network-policy report --input ./networktrace.log > network-policy-default.yaml

Prometheus & Grafana & eBPF exporter 設定

 Prometheus


編輯 prometheus.yml
- job_name: "eBPF"
static_configs:
- targets: ['localhost:9435']

執行Prometheus, Grafana and ebpf_exporter
~/git/ebpf_exporter$ sudo ~/SourceCode/golang/bin/ebpf_exporter --config.file=examples/bio.yaml

cd ~/Prometheus/prometheus-2.16.0.linux-amd64
./prometheus --config.file=prometheus.yml

cd ~/Downloads/grafana-6.0.0
./bin/grafana-server web

# admin/bbyanlfe
#http://140.96.27.24:3000/dashboard/

eBPF exporter 使用說明
build eBPF exporter and run it
cd ~/SourceCode/go_SourceCode/ebpf_exporter
go build cmd/ebpf_exporter/main.go
sudo ./main --config.file=examples/timers.yaml --debbug

#http://140.96.27.24:9435/metrics

visual studio code to run ebpf_exporter with sudo
sudo code --user-data-dir=/home/liudanny/
### launch.json ###
{
// Use IntelliSense to learn about possible attributes.
// Hover to view descriptions of existing attributes.
// For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
"version": "0.2.0",
"configurations": [
{
"name": "Launch",
"type": "go",
"request": "launch",
"mode": "auto",
"program": "${workspaceFolder}/cmd/ebpf_exporter/main.go",
"env": {},
"args": ["--config.file=${workspaceFolder}/examples/timers.yaml"]
}
]
}

[Cilium] Hubble Source Code Study

 

Source Code Study:

https://github.com/cilium/cilium/blob/master/bpf/bpf_lxc.c#L1223
int handle_policy(struct __ctx_buff *ctx)
-> tail_ipv4_policy()
    |
    -> ipv4_policy()
        |
        V
        https://github.com/cilium/cilium/blob/master/bpf/lib/policy_log.h#L46
        send_policy_verdict_notify()
在此function放入 struct policy_verdict_notify 格式化的資料
        |
    -> ctx_event_output()
   (這function是被remap來自skb_event_output()
   會將event送到perf's ring buffer in Kernel)
  
#Cilium 有 perf reader 可以從 Kernel 讀 perf's ring buffer 然後解開msg
https://github.com/cilium/ebpf/blob/master/perf/reader.go


#另外一個會送send_trace_notify()的來源 (有很多bpf的點都會送)
https://github.com/cilium/cilium/blob/master/bpf/bpf_network.c
int from_network(struct __ctx_buff *ctx)
   |
   V
   https://github.com/cilium/cilium/blob/master/bpf/lib/trace.h
   send_trace_notify()
   這個function會發送packet tracing events是很重要的flow來源

#上面兩個例子的notify type 是 CILIUM_NOTIFY_POLICY_VERDICT & CILIUM_NOTIFY_TRACE
#到了Hubble就會是Flow (Layer3/4)主要來源
case monitorAPI.MessageTypeDrop,
monitorAPI.MessageTypeTrace,
monitorAPI.MessageTypePolicyVerdict:
return p.l34.Decode(payload, decoded)
case monitorAPI.MessageTypeAccessLog:
return p.l7.Decode(payload, decoded)

 
BPF datapath events 定義在這
https://github.com/cilium/cilium/blob/master/pkg/monitor/api/types.go


#跟container有關的send_trace_notify()的來源
https://github.com/cilium/cilium/blob/master/bpf/bpf_lxc.c
tail_handle_ipv4()
-> handle_ipv4_from_lxc()
   |
   V
   https://github.com/cilium/cilium/blob/master/bpf/lib/trace.h
   send_trace_notify()
../_images/hubble_getflows.png

測試:

kubectl get pods -n kube-system

kubectl exec -n kube-system ds/cilium -- cilium status
hubble observe --verdict DROPPED --pod XXX
hubble observe --since 3m --protocol http --output=json | \
jq -sr 'map(select(.l7.type=="RESPONSE" and .l7.http.method=="GET")) |
group_by(.l7.http.url) |
map({url: .[0].l7.http.url, max_latency_ms: ((map(.l7.latency_ns | tonumber) | max) / 1000000 }) |
sort_by(.max_latency_ms)'

hubble observe --since 3m --protocol tcp --output=json | \
jq -sr 'map(select(.l4.TCP.destination_port==8080)) |
group_by(.l4.TCP.flags)'

hubble observe --since=10m -t l7 -j \
| jq 'select(.l7.dns.rcode==3) | .destination.namespace + "/" + .destination.pod_name' \
| sort | uniq -c | sort -r

hubble-relay serve --peer-service=unix:///var/run/cilium/hubble.sock --listen-address=:4245

curl \
--silent \
--unix-socket /var/run/cilium/hubble.sock \
--request GET \
http://localhost/ \
| jq .


hubble serve --listen-client-urls=0.0.0.0:50051 --listen-client-urls=unix:///var/run/hubble.sock --metrics-server :6943 --metric=dns:query --metric=drop --metric=tcp --metric=flow --

[ODL] Import ODL Controller using Maven into Eclipse on Windows 10

 

Prerequisites

Java

目前使用的是OpenJDK, 需在Windows的系統環境變數做"JAVA_HOME"設定, 變數值為安裝的JDK路徑

Eclipse

需先在Windows 10上安裝Eclipse IDE, 目前使用的版本是 2019-09 R (4.13)

Maven

下載並解壓縮完成後, 請在系統環境變數做"MAVEN_HOME" and "M2_HOME"設定如下:

Import ODL Controller using Maven into Eclipse

Eclipse 環境設定 for Maven路徑


Install m2e plugin

Help --> Install New Software
在 "Work with" 選擇 m2e release repository 或是 自行增加 ( 輸入 URL 後按下 "Add" )
勾選Maven Integration for Eclipse, 然後按下 "Finish"

把Maven's "Plugin execution not covered by lifecycle configuration" 改選為 "Ignore"

Import ODL Controller


切換到 tag: v3.0.2
$git checkout tags/v3.0.2

Project --> Import

以上述方式import project, 只剩下這種的Maven Error:
"Execution generate-depends-file of goal org.apache.servicemix.tooling:depends-maven-plugin:1.4.0:generate-depends-file failed"

Wednesday, September 15, 2021

[eBPF] The example of using BCC's trace.py script


If you want to use a container to run BCC's script, you can follow the instructions to build a Docker image and run it as a container.

Dockerfile

FROM ubuntu:18.04

RUN apt update && apt install -y lsb-core vim curl cscope cmake ctags file git locales bison flex iperf netperf android-tools-adb build-essential libedit-dev zlib1g-dev libelf-dev tree wget openjdk-8-jdk libgtk-3-dev iputils-ping
RUN apt-key adv --keyserver keyserver.ubuntu.com --recv-keys 4052245BD4284CDD
RUN echo "deb https://repo.iovisor.org/apt/$(lsb_release -cs) $(lsb_release -cs) main" > /etc/apt/sources.list.d/iovisor.list
RUN apt update && apt-get install -y bcc-tools libbcc-examples

Build a Docker image
sudo docker build -t bcc-ebpf .

Run the Docker image as a container 
sudo docker run -d --name bcc \
    --privileged \
    -v $(pwd):/bcc \
    -v /lib/modules:/lib/modules:ro \
    -v /usr/src:/usr/src:ro \
    -v /boot:/boot:ro \
    -v /sys/kernel/debug:/sys/kernel/debug \
    bcc-ebpf sleep 3600d

The examples of using BCC's trace.py
python trace.py 'r:bash:readline "%s", retval'
python trace.py 'u:/lib/x86_64-linux-gnu/libc-2.27.so:memory_sbrk_more "%u", arg1' -T

./trace.py 'python:_PyImport_LoadDynamicModule "name: %s path: %s" arg1, arg2' \
        'r:python:_PyImport_LoadDynamicModule "at 0x%x" retval'

./trace.py 'r:bash:readline "%s" retval'

./trace.py '/home/bgregg/functions:main.add "%d %d" arg1, arg2'

# List functions in the executable file
readelf -s test
objdump -t test
objdump -D test


Reference:
Linux eBPF/bcc uprobes
http://www.brendangregg.com/blog/2016-02-08/linux-ebpf-bcc-uprobes.html
Using user-space tracepoints with BPF

[eBPF] An example of Userspace Tracing

#!/usr/bin/python

from __future__ import print_function
from bcc import BPF
from time import sleep

# load BPF program

b = BPF(text="""
#include <uapi/linux/ptrace.h>
struct key_t {
    char c[80];
};
BPF_HASH(counts, struct key_t);

int count(struct pt_regs *ctx) {
    if (!PT_REGS_PARM1(ctx))
        return 0;

    struct key_t key = {};
    u64 zero = 0, *val;
    bpf_probe_read(&key.c, sizeof(key.c), (void *)PT_REGS_PARM1(ctx));
    val = counts.lookup_or_init(&key, &zero);
    (*val)++;
    return 0;
};
""")

b.attach_uprobe(name="c", sym="strlen", fn_name="count")

# header
print("Tracing strlen()... Hit Ctrl-C to end.")

# sleep until Ctrl-C
try:
    sleep(99999999)
except KeyboardInterrupt:
    pass

# print output
print("%10s %s" % ("COUNT", "STRING"))
counts = b.get_table("counts")
for k, v in sorted(counts.items(), key=lambda counts: counts[1].value):
    print("%10d \"%s\"" % (v.value, k.c.encode('string-escape')))


Reference:
https://github.com/iovisor/bcc/blob/master/tools/bashreadline.py
https://github.com/iovisor/bcc/blob/master/tools/gethostlatency.py
https://github.com/iovisor/bcc/blob/master/tools/funccount.py
https://github.com/iovisor/bcc/blob/master/tools/memleak.py
https://github.com/iovisor/bcc/blob/master/tools/dbslower.py
https://github.com/iovisor/bcc/blob/master/tools/trace.py


Friday, February 21, 2020

[Tracing] Use BCC tools to do dynamic tracing in Linux user space

Here is a very simple example of using BCC to trace a program and get the data from the function's argument.

A Simple Test Code in C: "test.c"
#include <stdio.h>

// student structure
struct student {
  char id[15];
  char firstname[64];
  char lastname[64];
  float points;
};
// function declaration
void getDetail(struct student *);
void displayDetail(struct student *);

int main(void) {
  // student structure variable
  struct student std[1];
  // get student detail
  getDetail(std);
  // display student detail
  displayDetail(std);
  return 0;
}

// function definition
void getDetail(struct student *ptr) {
  int i;
  for (i = 0; i < 1; i++) {
    printf("Enter detail of student #%d\n", (i + 1));
    printf("Enter ID: ");
    scanf("%s", ptr->id);
    printf("Enter first name: ");
    scanf("%s", ptr->firstname);
    printf("Enter last name: ");
    scanf("%s", ptr->lastname);
    printf("Enter Points: ");
    scanf("%f", &ptr->points);
    // update pointer to point at next element
    // of the array std
    ptr++;
  }
}

void displayDetail(struct student *ptr) {
  int i;
  for (i = 0; i < 1; i++) {
    printf("\nDetail of student #%d\n", (i + 1));
    // display result via ptr variable
    printf("\nResult via ptr\n");
    printf("ID: %s\n", ptr->id);
    printf("First Name: %s\n", ptr->firstname);
    printf("Last Name: %s\n", ptr->lastname);
    printf("Points: %f\n", ptr->points);
    // update pointer to point at next element
    // of the array std
    ptr++;
  }
}

The simple BCC example code: "mytrace.py"
It will get and print out the string of firstname when it runs
from __future__ import print_function
from bcc import BPF
from time import strftime
import argparse


# load BPF program
bpf_text = """
#include <uapi/linux/ptrace.h>
#include <linux/sched.h>
struct str_t {
    u64 pid;
    char str[80];
};

// student structure
struct student {
  char id[15];
  char firstname[64];
  char lastname[64];
  float points;
};

BPF_PERF_OUTPUT(events);
int printret(struct pt_regs *ctx) {
    struct str_t data  = {};
    char comm[TASK_COMM_LEN] = {};
    u32 pid;
    if (!PT_REGS_RC(ctx))
        return 0;
    pid = bpf_get_current_pid_tgid();
    data.pid = pid;
    bpf_probe_read(&data.str, sizeof(data.str),
        ((struct student *)PT_REGS_RC(ctx))->firstname);
    bpf_get_current_comm(&comm, sizeof(comm));
    events.perf_submit(ctx,&data,sizeof(data));
    return 0;
};
"""

b = BPF(text=bpf_text)
b.attach_uprobe(name="<...your test binary file location...>/test",
        sym="displayDetail", fn_name="printret")

# header
print("%-9s %-6s %s" % ("TIME", "PID", "ARGUMENT"))

def print_event(cpu, data, size):
    event = b["events"].event(data)
    print("%-9s %-6d %s" % (strftime("%H:%M:%S"), event.pid,
                            event.str.decode('utf-8', 'replace')))

b["events"].open_perf_buffer(print_event)
while 1:
    try:
        b.perf_buffer_poll()
    except KeyboardInterrupt:
        exit()
Here is the steps:
$ sudo python mytrace.py

# open another terminal
$ gcc -o test test.c
$ ./test
Enter detail of student #1
Enter ID: 1234
Enter first name: Danny
Enter last name: Liu
Enter Points: 100

Detail of student #1

Result via ptr
ID: 1234
First Name: Danny
Last Name: Liu
Points: 100.000000
Then you will see the tracing message in your previous terminal as follows:
(print out the first name from function's argument)
$ sudo python mytrace.py
TIME      PID    ARGUMENT
17:01:32  7265   Danny

Thursday, February 20, 2020

Thursday, February 13, 2020

[Darknet][Python] Support using Numpy array as image input instead of image path in Darknet Python API

When I dealt with Darknet's detect function in Python, I found that there is only one way for accepting an image file path as the argument, which is convenient for inferencing images from files.  But, it is not very well for images from video files or cameras.
Due to this reason, I dig out some solutions already on the Internet, and modify as follows:

Thursday, October 24, 2019

[Spring Boot] My first Spring Boot Project for demo

Recently, the term "Microservices" is very hot in the cloud service/architecture and it gets my attention. So, I want to learn more about what it is. Furthermore, I also notice Spring Cloud is a very popular microservice framework based on Spring Boot and it provides tools for developers to quickly build some of the common patterns in distributed systems (e.g. configuration management, service discovery, circuit breakers, intelligent routing, micro-proxy, control bus, one-time tokens, global locks, leadership election, distributed sessions, cluster state)

Monday, October 7, 2019

[Dockerfile] Some of the little skills used in Dockerfile

I collect some of the little skills used in my Dockerfiles and also keep in the record for my reference.

Thursday, September 19, 2019

[IJava] A Jupyter kernel for executing Java code

As we know that Jupyter’s is a Python-based solution, and it provides a lot of convenience for interactivity with Python code. But, I just wonder if Jupyter can use Java as its kernel, and it can. I find that it allows to integrate additional "kernels" with Java capabilities and someone has done it -IJava: https://github.com/SpencerPark/IJava

[Kubernetes] The simple introduction of VPA

VPA stands for Vertical Pod Autoscaling, which frees you from having to think about what values to specify for a container's CPU and memory requests. The autoscaler can recommend values for CPU and memory requests, or it can automatically update values for CPU and memory requests.

Before using VPA, we need to install Metrics Server first as follows:

Wednesday, August 21, 2019

[Kubernetes] Call Kubernetes APIs

If you want to use Bash Shell  to call Kubernetes APIs, Here is the way to get the token and use it as follows:

Prepare variables
#APISERVER="https://10.10.0.100:6443"
$ APISERVER=$(kubectl config view | grep server | cut -f 2- -d ":" | tr -d " ")
$ TOKEN=$(kubectl describe secret $(kubectl get secrets | grep default | cut -f1 -d ' ') | \
    grep -E '^token' | cut -f2 -d':' | tr -d '\t')
Call APIs
$ curl $APISERVER/api --header "Authorization: Bearer ${TOKEN//[[:space:]]/}" --insecure
{
  "kind": "APIVersions",
  "versions": [
    "v1"
  ],
  "serverAddressByClientCIDRs": [
    {
      "clientCIDR": "0.0.0.0/0",
      "serverAddress": "192.168.0.100:6443"
    }
  ]
}

[Docker] Troubleshooting to docker private registry

I create a private docker registry as follows and sometimes it cannot reply the http request.
$ sudo docker run -p 7443:7443 --restart=always \
  -v /raid/registry2:/var/lib/registry \
  -e REGISTRY_HTTP_ADDR=0.0.0.0:7443 \
  --name registry registry:2

$ curl -v http://192.168.10.10:7443/v2/_catalog
*   Trying 192.168.10.10...
* TCP_NODELAY set
* Connected to 192.168.10.10 (192.168.10.10) port 7443 (#0)
> GET /v2/_catalog HTTP/1.1
> Host: 192.168.0.109:7443
> User-Agent: curl/7.58.0
> Accept: */*
...(hang)...

Tuesday, August 20, 2019

[Kubernetes] Add new DNS server in CoreDNS Configuration

After finished the K8S installation, I had encountered an issue that the pod cannot resolve the domain name. I use these commands to check the DNS issue.
$ systemd-resolve --status
nameservers:
                  addresses:
                  - 210.240.232.1
                  search: []

# use busybox pod to run nslookup
$ kubectl apply -f https://k8s.io/examples/admin/dns/busybox.yaml
$ kubectl exec -it busybox -- nslookup kubernetes.default
$ kubectl exec -it busybox -- nslookup google.com

# check local DNS configuration
$ kubectl exec busybox cat /etc/resolv.conf

Monday, August 5, 2019

[Qt] How to install Qt for WebAssembly

In the previous post: [Qt] The Qt features of WebAssembly and Qt quick WebGLStreaming, I mentioned about Qt for WebAssembly but didn't try it yet. Here, this post is to introduce how to install Qt for WebAssembly and give an example to try by the following steps:

Tuesday, July 30, 2019

[Kubernetes] How to install Kubernetes Dashboard and without invalid certification

When I follow the instructions from the official site: https://github.com/kubernetes/dashboard to install Kubernetes Dashboard, I encounter the problem that I cannot access the dashboard via my browser because the certificate is invalid. After figuring it out, Here is my approach to resolving it.

Sunday, July 21, 2019

[DDS] Install OpenSplice DCPS Python API on Raspberry Pi 3

Before starting to introduce how to install OpenSplice DCPS Python API on Raspberry Pi 3, we can take a look at the guide of OpenSplice DCPS Python API as list:

Thursday, July 18, 2019

[Kubernetes] The example of commands for commonly checking kubernetes status and troubleshooting

This post is about the example of commands for checking kubernetes status and troubleshooting. The purpose is for the reference by myself.

Wednesday, July 17, 2019

[Python] The issues of converting Python2 to Python3

If you are working on converting Python2 source code to Python3, there are some issues you will encounter sooner or later. I arrange my part here and will continue to update it as follows:

Monday, July 15, 2019

[Qt] The Qt features of WebAssembly and Qt quick WebGLStreaming

I recently noticed that there are 2 major Qt features: Qt WebAssembly and Qt quick WebGL which are very powerful and useful. For the Qt WebAccembly, any your C++ Qt applications could be executed on a browser. For Qt quick WebGL, it is optimized for Qt Quick and allows you to run remote Qt Quick applications in a browser.

Qt WebAssembly

Qt for WebAssembly

Get started with Qt for WebAssembly
Here is a presentation about Qt for WebAccembly.
QtWS18 – Qt for WebAssembly by Morten Sørvig, The Qt Company

Qt WebAssembly Examples
https://github.com/msorvig/qt-webassembly-examples

Qt For WebAssembly Examples
(source code: https://github.com/msorvig/qt-webassembly-examples)
SensorTagDemo
colordebugger
gui_lifecycle
gui_localfiles
gui_opengl
gui_raster
mqtt_simpleclient
quick_clocks
quick_controls2_gallery
quick_controls2_testbench
quick_hellosquare
slate
widget_wiggly
widgets_wiggly

https://www.jianshu.com/p/bed6c2963435

Qt quick WebGL 

Here is a quick introduction to Qt quick WebGL

The introduction of Qt Quick WebGL Streaming

WebGL streaming in a Raspberry PI Zero W

Related Demo:

[Shared Library] How to check the used shared libraries in your program

This post is only to show the command about checking the used shared libraries in your program.

Wednesday, June 19, 2019

[DDS] Install OpenSplice Community Version

What is DDS? The Data Distribution Service (DDS™) is a middleware protocol and API standard for data-centric connectivity from the Object Management Group® (OMG®). It integrates the components of a system together, providing low-latency data connectivity, extreme reliability, and a scalable architecture that business and mission-critical Internet of Things (IoT) applications need. For more information, check it out: https://www.dds-foundation.org/what-is-dds-3/
https://zhuanlan.zhihu.com/p/32278571

In this post, I install OpenSplice as my DDS runtime environment and library. You can download it from here: https://www.adlinktech.com/en/dds-community-software-evaluation.aspx

Wednesday, June 12, 2019

[TensorFlow] Build TensorFlow from source with Intel MKL enabled

Based on this document "Intel® Math Kernel Library for Deep Learning Networks: Part 1–Overview and Installation",  I give a quick summary to do it.
Here are the steps for building TensorFlow from source with Intel MKL enabled.

For MKL DNN:
# Install Intel MKL & MKL-DNN
$ git clone https://github.com/01org/mkl-dnn.git
$ cd mkl-dnn
$ cd scripts && ./prepare_mkl.sh && cd ..
$ mkdir -p build && cd build && cmake .. && make -j$(nproc)
$ make test
$ sudo make install
For building TensorFlow:

Monday, June 10, 2019

[Squid] Setup Linux Proxy Server and Client

The following steps are about installing and setup Linux proxy server (Squid) 

#install squid proxy
$ sudo apt-get install squid

#modify squid.conf
$ sudo vi /etc/squid/squid.conf
  ==> #add your local network segment
  acl mynetwork src 192.168.0.0/24
  ==> #allow "mynetwork" be accessing via http
  http_access allow mynetwork

#restart squid...OK
$ sudo service squid restart

Tuesday, May 28, 2019

[Docker] Using GUI with Docker

Recently I need to run my GUI application with Docker and it can show up either directly on my desktop operating system or in my ssh terminal client via X11.

Basically, there are some people who already provide the solution for the cases. I just list the reference and quickly give my examples.

Monday, April 22, 2019

[TVM] Deploy TVM Module using C++ API

When the first time to deal with deploying TVM module using C++ API, I found this official web site: Deploy TVM Module using C++ API which only gives a fine example for deploying, but it doesn't explain how to generate the related files and modules.

So, after several times of trial and error, I figure it out to generate the required files for deploying.
Basically you can refer to the TVM tutorials about compiling models:
https://docs.tvm.ai/tutorials/index.html

Monday, April 15, 2019

[Experiment] Compare the inference performance of TensorFlow Lite and TVM

I compare the inference performance of both TensorFlow Lite and TVM on my laptop with the same MobileNet model and the same input size of 224*224.
They both assign two threads to do the inference task and see the average inferencing time it spent.
(P.S: Giving the same of 10 threads in these 2 cases )

Wednesday, April 10, 2019

[TensorFlow Lite] The performance experiment of TensorFlow Lite

Based on my previous post: [TensorFlow Lite] My first try with TensorFlow Lite ,
I am just curious about the performance of TensorFlow Lite on the different platforms so that I do a performance experiment for it.
There are a X86 server (no GPU enabled) and a Nvidia TX2 board ( which is ARM based ) to do the experiment. They both run TFLite’s official example using Inception_v3 model and get average inference time as the following table.

Hardware Information:
Nvidia TX2: 4 CPU cores and 1 GPU ( but does not use )
X86 server: 24 CPU cores and no GPU enabled

[XLA] Build Tensorflow XLA AOT Shared Library

Build Tensorflow XLA AOT Shared Library:

Add the following code into tensorflow/compiler/aot/BUILD in TensorFlow source:

cc_binary(
    name = "libxlaaot.so",
    deps = [":embedded_protocol_buffers",
        "//tensorflow/compiler/tf2xla:xla_jit_compiled_cpu_function",
        "//tensorflow/compiler/tf2xla",
        "//tensorflow/compiler/tf2xla:cpu_function_runtime",
        "//tensorflow/compiler/tf2xla:tf2xla_proto",
        "//tensorflow/compiler/tf2xla:tf2xla_util",
        "//tensorflow/compiler/tf2xla:xla_compiler",
        "//tensorflow/compiler/tf2xla/kernels:xla_cpu_only_ops",
        "//tensorflow/compiler/tf2xla/kernels:xla_dummy_ops",
        "//tensorflow/compiler/tf2xla/kernels:xla_ops",
        "//tensorflow/compiler/xla:shape_util",
        "//tensorflow/compiler/xla:statusor",
        "//tensorflow/compiler/xla:util",
        "//tensorflow/compiler/xla:xla_data_proto",
        "//tensorflow/compiler/xla/client:client_library",
        "//tensorflow/compiler/xla/client:compile_only_client",
        "//tensorflow/compiler/xla/client:xla_computation",
        "//tensorflow/compiler/xla/service:compiler",
        "//tensorflow/compiler/xla/service/cpu:buffer_info_util",
        "//tensorflow/compiler/xla/service/cpu:cpu_compiler",
        "//tensorflow/core:core_cpu_internal",
        "//tensorflow/core:framework_internal",
        "//tensorflow/core:lib",
        "//tensorflow/core:lib_internal",
        "//tensorflow/core:protos_all_cc",
        ":tfcompile_lib",
        "//tensorflow/compiler/xla/legacy_flags:debug_options_flags",
        "//tensorflow/core:core_cpu",
        "//tensorflow/core:framework",
        "//tensorflow/core:graph",
        "@com_google_absl//absl/memory",
        "@com_google_absl//absl/strings",
        "@com_google_absl//absl/types:span",
    ],
    linkopts=["-shared -Wl,--whole-archive" ],
    linkshared=1

)

Wednesday, March 27, 2019

[TensorFlow Lite] Build Tensorflow Lite C++ shared library

Build Tensorflow Lite shared library:

Modify the code: tensorflow/contrib/lite/BUILD in TensorFlow source:
cc_binary(
    name = "libtflite.so",
    deps = [":framework",
        "//tensorflow/contrib/lite/kernels:builtin_ops",
        "//tensorflow/contrib/lite/kernels:eigen_support",
        "//tensorflow/contrib/lite/kernels:gemm_support",
    ],
    linkopts=["-shared -Wl,--whole-archive" ],
    linkshared=1

)

[TFCompile] Use XLA AOT Compiler to compile Resnet50 model and do inference

I inspired by this following article and try to do something different because it's approach using Keras has an issue for XLA AOT compiler.
Kerasモデルをtfcompileでコンパイルする
Instead, I download the pre-trained Resnet50 model and optimize simply by the tool:transform_graph

Download:
wget http://download.tensorflow.org/models/official/20181001_resnet/savedmodels/resnet_v2_fp32_savedmodel_NHWC.tar.gz
or
wget http://download.tensorflow.org/models/official/20181001_resnet/savedmodels/resnet_v2_fp32_savedmodel_NHWC_jpg.tar.gz
Transform:
bazel-bin/tensorflow/tools/graph_transforms/transform_graph \
--in_graph='/home/liudanny/git/tensorflow_danny/tensorflow/compiler/aot/myaot_resnet50/resnetv2_imagenet_frozen_graph.pb' \
--out_graph='/home/liudanny/workspace/pyutillib/my_resnet50/optimized_frozen_graph.pb' \
--inputs='input_tensor:0' \
--outputs='softmax_tensor:0' \
--transforms='
  strip_unused_nodes
  fold_constants'

#I don't enable the follwoing options due to worse performance.
#fold_batch_norms      <== For XLA AOT, it is not good in the performance
#fold_old_batch_norms  <== For XLA AOT, it is not good in the performance
#quantize_weights' <== XLA AOT doesn't support

Thursday, March 21, 2019

[AutoKeras] My first try with a simple example of AutoKeras

AutoKeras only supports Python 3.6 so that the running environment has to install Python 3.6. My operation system is Ubuntu 16.04 and it needs to add apt repository first.

Install Python 3.6 and AutoKeras ( Don't remove Python 3.5)
# Install pip3
apt-get install python3-pip
# Install Python 3.6
apt-get install software-properties-common
add-apt-repository ppa:jonathonf/python-3.6
apt-get update
apt-get install python3.6

update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.5 1
update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.6 2
update-alternatives --config python3

ln -s /usr/include/python3.5 /usr/include/python3.6m
pip3 install lws
pip3 install autokeras

Friday, March 15, 2019

[TensorFlow] Build TensorFlow v1.12 from the source on Ubuntu 16.04

My previous post: [TensorFlow] How to build your C++ program or application with TensorFlow library using CMake
It is for building TensorFlow from the source based on v1.10. Currently, I want to upgrade it to v1.12 and encounter some problems. 
First, my version of ProtoBuf library on my system is v3.6.1 so that we should align its version in the TensorFlow.
Second, it seems that there are a few issues when building TensorFlow v1.12 that we need to deal with it case by case.

Monday, March 11, 2019

[TensorFlow Lite] My first try with TensorFlow Lite

I just take my first try with the example: label_image (tensorflow/contrib/lite/examples/label_image) in TensorFlow Lite and write down the commands that I used.
There are a bunch of information from the offical TensorFlow Lite guide:
https://www.tensorflow.org/lite/guide

1. convert the example of model to tflite format

Wednesday, March 6, 2019

[Tool] Convert TensorFlow graph to UFF format

The previous post: How to use TensorRT to do inference with TensorFlow model ? has introduced away to do the converting job for UFF format model. But, basically there are 2 ways to do that:

1. Convert TensorFlow's Session GraphDef directly on the fly to UFF format model
    ==> convert_uff_from_tensorflow()
2. Convert the frozen model file to UFF format model
    ==> convert_uff_from_frozen_model()

The following code is about the functions to convert TensorFlow graph to UFF format for running with TensorRT.

Tuesday, March 5, 2019

[OpenCV] Build OpenCV 3.4.4 on TX2

For the reason that I was curious about the performance of OpenCV on TX2 using GPU, I installed OpenCV 3.4.4 (this version and after will integrate with the inference engine) on my TX2 based on the following links.
https://jkjung-avt.github.io/opencv3-on-tx2/
https://www.learnopencv.com/install-opencv-3-4-4-on-ubuntu-16-04/

Monday, February 25, 2019

[Inspecting Graphs] Use TensorFlow's summarize_graph tool to find the input and output node names in the frozen model/graph

When trying to do inferencing using a frozen model from downloading or freezing by yourself, we may encounter a problem about what the input and output node names are in this model? If we cannot figure them out, it is impossible for you to do inferencing correctly. Here is an easy way to get the possible ones: using the tool: "summarize_graph"

bazel build tensorflow/tools/graph_transforms:summarize_graph
bazel-bin/tensorflow/tools/graph_transforms/summarize_graph --in_graph=/danny/tmp/faster_rcnn_resnet101_coco_2018_01_28/frozen_inference_graph.pb
Found 1 possible inputs: (name=image_tensor, type=uint8(4), shape=[?,?,?,3])
No variables spotted.
Found 4 possible outputs: (name=detection_boxes, op=Identity) (name=detection_scores, op=Identity) (name=num_detections, op=Identity) (name=detection_classes, op=Identity)
Found 48132698 (48.13M) const parameters, 0 (0) variable parameters, and 4163 control_edges
Op types used: 4688 Const, 885 StridedSlice, 559 Gather, 485 Mul, 472 Sub, 462 Minimum, 369 Maximum, 304 Reshape, 276 Split, 205 RealDiv, 204 Pack, 202 ConcatV2, 201 Cast, 188 Greater, 183 Where, 149 Shape, 145 Add, 109 BiasAdd, 107 Conv2D, 106 Slice, 100 Relu, 99 Unpack, 97 Squeeze, 94 ZerosLike, 91 NonMaxSuppressionV2, 55 Enter, 46 Identity, 45 Switch, 27 Range, 24 Merge, 22 TensorArrayV3, 17 ExpandDims, 15 NextIteration, 12 TensorArrayScatterV3, 12 TensorArrayReadV3, 10 TensorArrayWriteV3, 10 Exit, 10 Tile, 10 TensorArrayGatherV3, 10 TensorArraySizeV3, 6 Transpose, 6 Fill, 6 Assert, 5 Less, 5 LoopCond, 5 Equal, 4 Round, 4 Exp, 4 MaxPool, 3 Pad, 2 Softmax, 2 Size, 2 GreaterEqual, 2 TopKV2, 2 MatMul, 1 All, 1 CropAndResize, 1 ResizeBilinear, 1 Relu6, 1 Placeholder, 1 LogicalAnd, 1 Max, 1 Mean
To use with tensorflow/tools/benchmark:benchmark_model try these arguments:
bazel run tensorflow/tools/benchmark:benchmark_model -- --graph=/danny/tmp/faster_rcnn_resnet101_coco_2018_01_28/frozen_inference_graph.pb --show_flops --input_layer=image_tensor --input_layer_type=uint8 --input_layer_shape=-1,-1,-1,3 --output_layer=detection_boxes,detection_scores,num_detections,detection_classes

For more information, please refer to this:
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/tools/graph_transforms#inspecting-graphs

Wednesday, January 30, 2019

[TFRecord] The easy way to verify your TFRecord file

There is a common situation when you build your TFRecord file ( your dataset ) and want to verify the correctness of the data in it. How to do it? I assume you don't have the problem to build your TFRecord file. So, the easy way to verify your TFRecord file is to use the API: tf.python_io.tf_record_iterator()

Wednesday, January 16, 2019

[TensorFlow] How to use Distribution Strategy in TensorFlow?

Learned from What’s coming in TensorFlow 2.0, TensorFlow 2.0 is coming soon and there are several features which are ready to use already, for instance, Distribution Strategy. Quoted from the article,
"For large ML training tasks, the Distribution Strategy API makes it easy to distribute and train models on different hardware configurations without changing the model definition. Since TensorFlow provides support for a range of hardware accelerators like CPUs, GPUs, and TPUs, you can enable training workloads to be distributed to single-node/multi-accelerator as well as multi-node/multi-accelerator configurations, including TPU Pods. Although this API supports a variety of cluster configurations, templates to deploy training on Kubernetes clusters in on-prem or cloud environments are provided."

Friday, January 11, 2019

[Shell] The example shell script of automation way to build the software or library required

I don't like to preach at people. But, if someone has a task that has been done more than two times, then he/she should consider using an automation way to ease the burden. One of the solutions is by writing a shell script. Here is an example of building OpenCV 3.4.1 on TX2 using a shell script. The software or the target platform is not the key part. The most important part is to adopt this sort of shell script to become the one of your version.

Thursday, January 10, 2019

[Tool] Visdom is a great visualization tool

I cannot use my word to describe Visdom because it is so amazingly awesome. Referencing the introduction from Facebook AI's official site,Visdom is a visualization tool that generates rich visualizations of live data to help researchers and developers stay on top of their scientific experiments that are run on remote servers. Visualizations in Visdom can be viewed in browsers and easily shared with others.

Monday, January 7, 2019

[TensorRT] How to use TensorRT to do inference with TensorFlow model ?

TensorRT is a high-performance deep learning inference optimizer and runtime that delivers low latency, high-throughput inference for deep learning applications. Here I am going to demonstrate that how to use TensorRT to do inference with TensorFlow model.


Install TensorRT
Please refer to this official website first:
https://docs.nvidia.com/deeplearning/sdk/tensorrt-install-guide/index.html#installing
After downloading TensorRT 4.0 ( in my case ), we can install it.
$ dpkg -i nv-tensorrt-repo-ubuntu1604-cuda9.0-ga-trt4.0.1.6-20180612_1-1_amd64.deb
$ apt-get update
$ apt-get install tensorrt
$ apt-get install python-libnvinfer-dev
$ apt-get install uff-converter-tf

Friday, January 4, 2019

[TensorFlow] How to write op with gradient in python?

Recently for some reasons, I studied the Domain-Adversarial Training of Neural Networks and it can be downloaded from http://jmlr.org/papers/volume17/15-239/15-239.pdf

In this paper, there is the key point that we should implement "Gradient Reversal Layer" for Discriminator to use it to connect the feature extractor. I found the source to implement it by replacing Identity op's gradient function as follows:

Thursday, January 3, 2019

[TensorFlow] How to generate the Memory Report from Grappler?

In the previous post, I introduce the way to generate cost and model report from Grappler.
https://danny270degree.blogspot.com/2019/01/tensorflow-how-to-generate-cost-and.html
In this post, I will continue to introduce the memory report which I think that is very useful. Please refer to my previous post to get the model code.

[TensorFlow] How to generate the Cost and Model Report from Grappler?

General speaking, Grappler in Tensorflow has several optimizers to do the specific area optimizations, such as for reducing the peak memory usage in GPU. So, I want to introduce some useful functions inside Grappler which are used for Simple Placer mechanism. And, these functions are also partially used in Grappler's optimizers.

Monday, December 24, 2018

[TensorFlow] My example of using SavedModelBuilder to do inference in TensorFlow

The purpose of this post is to show my example of SavedModelBuilder to do inference in TensorFlow. From my experiment, this approach can save a model with the signature that has input and output node name. And SavedModelBuilder can restore the graph based on the previously saved model pb file and the signature definition. Once, the restore is done, the inference task can be executed directly without GPU device needed if the training task is on GPU device.