Task Management

Procedure

1. Click next to System Profiler.

   Choose General Analysis from the drop-down list. The page for creating a task is displayed.

2. Set task parameters.
   

   • Click Import Template next to the Task Name text box to select and import a saved task template.
   • This section describes only common task parameters. For details about how to set other task parameters, see the corresponding analysis task sections.
   • The L3C access bandwidth and hit rate are not displayed on some OSs.
   • Some VMs (such as Huawei Cloud ECS) do not support the vPMU. As a result, the performance analysis task for collecting hardware PMU events will fail on these VMs.
   • When creating a task, do not use the following high-risk commands in the application parameter configuration items:

     cat, view, more, less, head, tail, awk, grep, egrep, fgrep, stty, rm, vi, sed, dos2unix, cp, mv, gzip, unzip, gunzip, tar, killall, pkill, chkconfig, mount, rpm, umount, fdisk, su, sudo, chmod, chown, chgrp, useradd, userdel, groupadd, groupdel, kill, shutdown, init, reboot, halt, poweroff, passwd, mkfs, ifconfig, ethtool, route, sh, bash, ksh, csh, crontab, sysctl, perf

   Table 1 Parameters for adding a task

   Parameter	Description
   Task Name	Name of the analysis task. A created task is automatically named in the Analysis type_Taskn format, with n starting from 1 in ascending order. You can change the name as required. The name must meet the following requirements: Contain only letters, digits, and underscores (_). Contain 1 to 64 characters.
   Select Nodes	Select the nodes to be analyzed. A maximum of 10 nodes are supported. You can search for nodes by node IP address or node name to filter nodes by status.
   Analysis Object	Object to be analyzed. The options are: System: Collects performance data of the server, regardless of the types of applications running in the system. The sampling duration is determined by the parameter specified. This option is recommended for servers running multiple types of services and subprocesses. Application: Collects data of a specified application or process for analysis. After selecting Application, select a mode for the analysis object. The options are: Launch application: Starts the application when the sampling begins. The sampling duration is determined by the execution time of the application. This option is recommended for applications with short execution time. The analysis covers the application and its subprocesses. Attach to process: Associates the existing process identifiers (PIDs) on the server during data collection. You need to configure the sampling duration parameter (applicable to scenarios where the application runs for a long time). The analysis covers only the PIDs in the parameter, not the sub-processes of the PIDs. NOTE: The parameters to be configured vary with the analysis type. For details, see the corresponding analysis task sections.
   Analysis Type	Type of the analysis task to be performed. The system analysis objects include: General analysis Overall analysis Hotspot functions System component analysis Microarchitecture Memory access I/O Specified analysis Resource scheduling HPC cluster check The application analysis objects include: General analysis Hotspot functions System component analysis Microarchitecture Memory access I/O Specified analysis Resource scheduling Locks and waits HPC application Roofline NOTE: The parameters to be configured vary with the analysis type. For details, see the corresponding analysis task sections. Resource scheduling, I/O analysis, memory access analysis, and roofline analysis tasks do not support a version upgrade. As a result, reports of earlier versions may not be displayed in the new version. Hotspot function analysis tasks for a single node have been adapted to Orange Pi Kunpeng Pro.
   Sampling Type	Type of sampled metrics.
   Sampling Duration	Task execution duration.

3. Click OK. For details about task parameters, see Table 1 and the related analysis task sections.

   You can click the icons next to the task name to perform the following operations:

   • : restarts the analysis task. You can modify task parameter settings and restart an analysis task. This button is available when a task is canceled or fails.
   • : cancels the analysis task. After an analysis task is canceled, the collected information will be deleted.
   • : deletes the analysis task. After a task is deleted, all data of this task will be deleted. Exercise caution when performing this operation.
   • : performs the analysis again. The analysis task is renamed and restarted.
   • : changes the task or report name. The report naming rule is the same as that of a task.
   

   • The administrator devadmin and common users can view, restart, and delete analysis tasks created by themselves. When creating a task, users can click Manage Nodes next to Select Nodes to go to the node management page.
   • If multiple tasks (including tasks of different users) are allocated to the same node, the tasks are executed in the order of submission time and only one task can be executed at a time.
4. (Optional) Cancel data collection or cancel the whole task.

   When the task is running, you can click Start Analysis to stop data collection and start analysis immediately or click Cancel to stop the task.

   

   This analysis function does not apply to HPC applications.

   Figure 1 Canceling a task
   
5. (Optional) Import or export a task.
   After an analysis task is created, you can hover the mouse cursor over the task name to view the task details. Analysis tasks can be imported and exported. See Figure 2.

   

   • Click or next to System Profiler. You can view and delete imported tasks, and download and delete exported tasks. If an import or export task fails, you can click Retry to import or export again.
   • The administrator can export tasks created by all users. Common users can export only tasks created by themselves. To view details, you can click in the upper right corner, choose Tool Settings > System Profiler > Import/Export Task.
   • If the name of the TAR package to be imported is changed, the original package name generated by the Kunpeng DevKit is displayed after the import.

   Figure 2 Importing/Exporting tasks
   

   • Click . A window shown in Figure 3 appears. Select either of the import modes as required.
     Figure 3 Importing a task
     
     

     • If Import Mode is set to Upload file, select the exported compressed file and upload it.
     • If Import Mode is set to Specify file path, enter the absolute path of the file on the server. The devkitworker1 user must have the read permission on the specified file as well as the access and write permissions on the directory where the file is located.
   • Click . A window shown in Figure 4 appears.
     Figure 4 Exporting a task
     
     

     You can collect data offline and then import the data for analysis.

     If the size of data to be exported exceeds 1 GB, the data is compressed into multiple compressed fragment sequence files whose size does not exceed 1 GB.

6. (Optional) Create a task template.

   Click in the upper right corner. On the page that is displayed, choose Tool Settings > System Profiler > Task Templates.

   

   • O the task creation page, you can click Save as Template and enter a template name to save the configuration information of the current task as a template.
   • The newly created template is displayed in the > Tool Settings > System Profiling > Task Templates area. You can view or (batch) delete task templates on the task template list page.
   • A maximum of 10 templates can be saved for each analysis task type. All users can only browse, import, and delete task templates created by themselves.
   Figure 5 Task template