Supported Process Metrics - Teradata Ecosystem Manager

Teradata Ecosystem Manager API Reference

Product
Teradata Ecosystem Manager
Release Number
16.10
Published
June 2017
Language
English (United States)
Last Update
2018-03-29
dita:mapPath
crq1488477755125.ditamap
dita:ditavalPath
ft:empty
dita:id
B035-3204
lifecycle
previous
Product Category
Analytical Ecosystem

The Ecosystem Manager Monitor supports the following three metrics:

Metric Description
PING Checks that a process is running at regular intervals. Teradata strongly recommends an interval of 300 seconds for all metrics.
Specifying a lower interval value causes significant overhead and affect overall performance.
MSM_USG Resident memory usage of a process.
CPU_USG Percentage CPU usage by a process.

User-specified input should mirror the following high-level XML structure:

<metricCollector monitorClass="com.teradata.tmsm.da.monitor.sigar.TDProcMonitor" interval="300" name="TDProcMon">
<metric resourceType=<METRIC_TYPE>[<base_process_name>] displayName=<process_display_name> >
   <arg>process_search_criteria</arg>
   <arg>process_search_criteria_value</arg>
</metric>
<reporter class="com.teradata.tmsm.da.monitor.reporter.TMSMPublisher"/>
</metricCollector>
          

where

  • METRIC_TYPE is one of the following: PING, MEM_USG, or CPU_USG. This is a required parameter. There can be more than one metric node (<metric>…</metric>) in the XML structure. For more information about correct usage, see the examples in this section.
  • base_process_name is the base process name of the process being monitored. This is a required parameter. These are the recommended values:
    • For C/C++-based executables, this should be the name of the executable.
    • For shell-based scripts, this should be the name of the shell (sh/bash).
    • For Java-based executables, this should be “java.”
  • process_display_name determines how the process name displays in the UI. This is an optional, but recommended parameter. Where multiple processes map to a process search criteria value, the display name is suffixed by an auto-generated sequence number.
  • process_search_criteria determines how the user wants to monitor a process. This is a required parameter. Use one of the process search criteria values listed in the following table.
    Value Function ps –ef Command Output XML Entries
    fullPathToProcExec Filters a process when the fully qualified path of the process executable you want to monitor is known /opt/sampleApp/sampleExe arg1 arg2 arg3
    <metric resourceType=”PING[sampleExe] displayName=”dispExe”>
    <arg>fullPathToProcExec</arg>
    <arg>/opt/sampleApp/sampleExe</arg></metric>
    commandLineArg

    Filters a process that is monitored using process command line arguments.

    Separate multiple process command line arguments with a space. For Microsoft Windows, use the tilde character (~) as the delimiter.

    /bin/sh ./sample.sh message 1000 localhost sampleLog1.lo
    <metric resourceType=”PING[sh] displayName=”sampleScript”>
    <arg>commandLineArg</arg><arg>./sample.sh message 1000 localhost sampleLog1.log</arg></metric>
    javaFullClassName Filters a java-based process that is initiated by specifying the full classname. /usr/bin/java –classpath /opt/sampleApp/sampleJavaApp.jar com.sampleCom.samplePackage.sampleJavaApp
    <metric resourceType=”PING[java] displayName=”sampleJavaApp”>
    <arg> javaFullClassName </arg>
    <arg com.sampleCom.samplePackage.sampleJavaApp </arg>
    </metric>
  • process_search_criteria_value is the value of the process search criteria. This value is case-sensitive and is shown in the XML entries column in the above table.
    If there is more than one process match for the user-specified process search criteria value, Ecosystem Manager Monitor reports back metrics for the multiple processes that match. If you expected or want only a single process to be monitored and reported, you must stop the emmonitor and edit the process search criteria value in emmonitorconfig.xml so that it maps to the single process that you want to monitor. Another way to view the output of a single process search criteria value is to issue a ps –ef command. However, this feature cannot report metrics for a process, if the native OS does not permit querying process metrics for various reasons, for example due to security policies.

Best Practices

The quickest way to determine the uniqueness of a process search criteria value is to execute the ps –ef command:

ps –ef | grep <process_search_criteria_value>

On Microsoft Windows, you can check the information in the Task-Manager or Services dialog box (path to executable). You can also refer to the emmonitor log file to find the log entries that relate the process_display_name and its PID. The search string for that log entry is Displayname .