Commit 7142dd73 authored by jinhai's avatar jinhai
Browse files

Merge branch 'fix_thermal_bug' into 'branch-0.5.0' 

MS-601 Docker logs error caused by get CPUTemperature error

See merge request megasearch/milvus!666

Former-commit-id: 5dfb1e644bd65d932cd42cb9cede082ff3ab9da9
parents a594ab9a 2cb6a5d1

cpp/CHANGELOG.md

+1 −0
Original line number Diff line number Diff line
@@ -10,6 +10,7 @@ Please mark all change in change log and use the ticket from JIRA. 
- MS-577 - Unittest Query randomly hung

- MS-587 - Count get wrong result after adding vectors and index built immediately

- MS-599 - search wrong result when table created with metric_type: IP

- MS-601 - Docker logs error caused by get CPUTemperature error



## Improvement

- MS-552 - Add and change the easylogging library

cpp/src/metrics/PrometheusMetrics.cpp

+1 −1
Original line number Diff line number Diff line
@@ -46,7 +46,7 @@ PrometheusMetrics::Init() { 
            return s.code();

        }



        const std::string uri = std::string("/tmp/metrics");

        const std::string uri = std::string("/metrics");

        const std::size_t num_threads = 2;



        // Init Exposer

cpp/src/metrics/SystemInfo.cpp

+39 −14
Original line number Diff line number Diff line
@@ -16,6 +16,7 @@ 
// under the License.



#include "metrics/SystemInfo.h"

#include "utils/Log.h"



#include <nvml.h>

#include <sys/types.h>
@@ -24,6 +25,9 @@ 
#include <iostream>

#include <string>

#include <utility>

#include<stdlib.h>

#include<dirent.h>

#include<stdio.h>



namespace milvus {

namespace server {
@@ -60,12 +64,12 @@ SystemInfo::Init() { 
    nvmlReturn_t nvmlresult;

    nvmlresult = nvmlInit();

    if (NVML_SUCCESS != nvmlresult) {

        printf("System information initilization failed");

        SERVER_LOG_ERROR <<  "System information initilization failed";

        return;

    }

    nvmlresult = nvmlDeviceGetCount(&num_device_);

    if (NVML_SUCCESS != nvmlresult) {

        printf("Unable to get devidce number");

        SERVER_LOG_ERROR << "Unable to get devidce number";

        return;

    }
@@ -151,7 +155,7 @@ SystemInfo::getTotalCpuTime(std::vector<uint64_t>& work_time_array) { 
    std::vector<uint64_t> total_time_array;

    FILE* file = fopen("/proc/stat", "r");

    if (file == NULL) {

        perror("Could not open stat file");

        SERVER_LOG_ERROR << "Could not open stat file";

        return total_time_array;

    }
@@ -162,7 +166,7 @@ SystemInfo::getTotalCpuTime(std::vector<uint64_t>& work_time_array) { 
        char buffer[1024];

        char* ret = fgets(buffer, sizeof(buffer) - 1, file);

        if (ret == NULL) {

            perror("Could not read stat file");

            SERVER_LOG_ERROR << "Could not read stat file";

            fclose(file);

            return total_time_array;

        }
@@ -237,19 +241,40 @@ SystemInfo::GPUTemperature() { 
std::vector<float>

SystemInfo::CPUTemperature() {

    std::vector<float> result;

    for (int i = 0; i <= num_physical_processors_; ++i) {

        std::string path = "/sys/class/thermal/thermal_zone" + std::to_string(i) + "/temp";

        FILE* file = fopen(path.data(), "r");

        if (file == nullptr) {

            perror("Could not open thermal file");

    std::string path = "/sys/class/hwmon/";



    DIR *dir = NULL;

    dir = opendir(path.c_str());

    if (!dir) {

        SERVER_LOG_ERROR << "Could not open hwmon directory";

        return result;

    }



    struct dirent *ptr = NULL;

    while ((ptr = readdir(dir)) != NULL) {

        std::string filename(path);

        filename.append(ptr->d_name);



        char buf[100];

        if (readlink(filename.c_str(), buf, 100) != -1) {

            std::string m(buf);

            if (m.find("coretemp") != std::string::npos) {

                std::string object = filename;

                object += "/temp1_input";

                FILE *file = fopen(object.c_str(), "r");

                if (file == nullptr) {

                    SERVER_LOG_ERROR << "Could not open temperature file"

                    exit(1);

                }

                float temp;

                fscanf(file, "%f", &temp);

                result.push_back(temp / 1000);

        fclose(file);

            }

        }

    }

    closedir(dir);

    return result;

}



std::vector<uint64_t>

SystemInfo::GPUMemoryUsed() {