Monitoring for Long I/O Metrics

When the time for disk I/Os to complete begins to lengthen for single or multiple devices, it may indicate potential storage disk problems such as maximum limit, degraded state (RAID set rebuilding), or fabric trouble. To take proactive measures to monitor the performance of disk I/O, follow the procedure below.

In this procedure you will run the batch file, DcsDiskPerfLog.bat, which defines the data collector sets DcsDiskPerfLog and DcsDiskPerfAlert. DcsDiskPerfAlert will monitor four of the DataCore Disk counters. If any counter exceeds the threshold, the system will start DcsDiskPerfLog, which will collect detailed performance statistics for the performance objects DataCoreDisk, LogicalDisk, PhysicalDisk, and DataCore SCSI Transport.

Instructions for Windows Server 2008:

  1. On the storage server that you want to monitor, run (“as administrator”) the batch file, DcsDiskPerfLog.bat.
    The batch file is located in %systemdrive%WindowsSysWOW64 (x64 systems).   

  2. In Computer Management, click System Tools>Reliability and Performance.

  3. In the Reliability and Performance Monitor, open Data Collector Sets>User Defined. You will see the two data collector sets DcsDiskPerfLog and DcsDiskPerfAlert in the console tree.
    SANmelody dcsdiskperflog1 Monitoring for Long I/O Metrics

  4. In the tree, select DcsDiskPerfLog and choose Properties from the shortcut menu.

  5. In the General tab, in the RunAs area, click Change….
    SANmelody dcsdiskperflog2 Monitoring for Long I/O Metrics

  6. In the box that opens, enter the administrator account and password used to run SANmelody software and click OK.
    SANmelody dcsdiskperflog3 Monitoring for Long I/O Metrics

  7. Click OK to close the properties box.

  8. Repeat the steps above for DcsDiskPerfAlert to assign the same administrator account and password in the Run As area.

  9. In the console tree, select DcsDiskPerfAlert and choose Start from the shortcut menu.
    SANmelody dcsdiskperflog4 Monitoring for Long I/O Metrics

  10. Now the system monitor is polling disk counters, looking for “long I/O’s.” If any counter exceeds the threshold, the system will start DcsDiskPerfLog, which will collect detailed performance statistics. Any long I/Os will produce .blg file in %systemdrive%PerfLogsAdmin.

NOTE:  You can customize the notification of alerts to run tasks if desired. This can be done from the Alert Task tab in the Properties box for DcsDiskPerfAlert, and from the Task tab in the Properties box for DcsDiskPerfLog. Refer to the Windows Help for more information.

Related topic:

DataCore Performance Objects.

 

Monitoring for Long I/O Metrics