Installing the GPU Metrics Collection Plug-in (Linux)

Scenarios

This topic describes how to install the plug-in to collect GPU and RAID metrics.

• ECSs support GPU metrics while BMSs do not.
• BMSs support RAID metrics while ECSs do not.
• If the Agent is upgraded to 1.0.5 or later, the corresponding plug-in must use the latest version. Otherwise, the metric collection will fail.

Prerequisites

• The Agent has been installed and is running properly.
• GPU metric collection requires ECSs to support GPU.
• Run the following command to check the Agent version:

  if [[ -f /usr/local/uniagent/extension/install/telescope/bin/telescope ]]; then /usr/local/uniagent/extension/install/telescope/bin/telescope -v; elif [[ -f /usr/local/telescope/bin/telescope ]]; then echo "old agent"; else echo 0; fi

  • If old agent is displayed, the early version of the Agent is used.
  • If a version is returned, the new version of the Agent is used.
  • If 0 is returned, the Agent is not installed.

Procedure (New Version)

1. Log in to an ECS as user root.
   

   • To monitor the BMS software RAID metrics, log in to a BMS.
   • The examples in the following procedure are based on the GPU plug-in installation. The installation for the software RAID plug-in is similar.
2. Run the following command to go to the Agent installation path /usr/local/telescope:

   cd /usr/local/uniagent/extension/install/telescope

3. Run the following command to create the plugins folder:

   mkdir plugins

4. Run the following command to enter the plugins folder:

   cd plugins

5. To download the script of the GPU metric collection plug-in, run the following command:

   wget https://telescope-eu-west-101.obs.eu-west-101.myhuaweicloud.eu/gpu_collector

   Table 1 Obtaining the plug-in installation package

   Name	Download Path
   Linux 64-bit installation package of the GPU metric collection plug-in	eu-west-101: https://telescope-eu-west-101.obs.eu-west-101.myhuaweicloud.eu/gpu_collector

6. Run the following command to add the script execution permissions:

   chmod 755 gpu_collector

7. Run the following command to create the conf.json file, add the configuration content, and configure the plug-in path and metric collection period crontime, which is measured in seconds:

   vi conf.json

   GPU metric plug-in configuration

   { 
      "plugins": [ 
        { 
          "path": "/usr/local/uniagent/extension/install/telescope/plugins/gpu_collector", 
          "crontime": 60 
        }
     ] 
   }

   RAID metric plug-in configuration

   { 
      "plugins": [ 
        { 
          "path": "/usr/local/uniagent/extension/install/telescope/plugins/raid_monitor.sh", 
          "crontime": 60 
        }
     ] 
   }

   

   • The parameters gpu_collector and raid_monitor.sh indicate the GPU plug-in and RAID plug-in configuration.
   • The collection period of the plug-in is 60 seconds. If the collection period is incorrectly configured, the metric collection will be abnormal.
   • Do not change the plug-in path without permission. Otherwise, the metric collection will be abnormal.
8. Open the conf_ces.json file in the /usr/local/uniagent/extension/install/telescope/bin directory. Add "EnablePlugin": true to the file to enable the plug-in to collect metric data.

   {    
       "Endpoint": "Region address. Retain the default value.",
       "EnablePlugin": true
   }

9. Restart the Agent:

   ps -ef | grep telescope | grep -v grep | awk '{print $2}' | xargs kill -9

Procedure (for the Early Version of the Agent)

1. Log in to an ECS as user root.
   

   • To monitor the BMS software RAID metrics, log in to a BMS.
   • The examples in the following procedure are based on the GPU plug-in installation. The installation for the software RAID plug-in is similar.
2. Run the following command to go to the Agent installation path /usr/local/telescope:

   cd /usr/local/telescope

3. Run the following command to create the plugins folder:

   mkdir plugins

4. Run the following command to enter the plugins folder:

   cd plugins

5. To download the script of the GPU metric collection plug-in, run the following command:

   wget https://telescope-eu-west-101.obs.eu-west-101.myhuaweicloud.eu/gpu_collector

   Table 2 Obtaining the plug-in installation package

   Name	Download Path
   Linux 64-bit installation package of the GPU metric collection plug-in	eu-west-101: https://telescope-eu-west-101.obs.eu-west-101.myhuaweicloud.eu/gpu_collector

6. Run the following command to add the script execution permissions:

   chmod 755 gpu_collector

7. Run the following command to create the conf.json file, add the configuration content, and configure the plug-in path and metric collection period crontime, which is measured in seconds:

   vi conf.json

   GPU metric plug-in configuration

   { 
      "plugins": [ 
        { 
          "path": "/usr/local/telescope/plugins/gpu_collector", 
          "crontime": 60 
        }
     ] 
   }

   RAID metric plug-in configuration

   { 
      "plugins": [ 
        { 
          "path": "/usr/local/telescope/plugins/raid_monitor.sh", 
          "crontime": 60 
        }
     ] 
   }

   

   • The parameters gpu_collector and raid_monitor.sh indicate the GPU plug-in and RAID plug-in configuration.
   • The collection period of the plug-in is 60 seconds. If the collection period is incorrectly configured, the metric collection will be abnormal.
   • Do not change the plug-in path without permission. Otherwise, the metric collection will be abnormal.
8. Open the conf_ces.json file in the /usr/local/telescope/bin directory. Add "EnablePlugin": true to the file to enable the plug-in to collect metric data.

   {    
       "Endpoint": "Region address. Retain the default value.",
       "EnablePlugin": true
   }

9. Run the following command to restart the Agent:

   /usr/local/telescope/telescoped restart