Tracking evaluation efficiency on endpoints with performance counters

Technote (troubleshooting)


Problem(Abstract)

The following article contains an analysis that can be activated for reporting on the efficiency at which an agent makes it through its evaluation cycle.

Symptom

Symptoms include when an agent takes a long time to gather content, process actions, and/or report results.


Resolving the problem

Use the attached analysis to track and have the agent report on the following performance counter data:

General Properties

    Version - The version of the agent
    Build Target - The type of machine the agent was compiled on
    Maximum cycle seconds - The amount of time (in seconds) the longest evaluation cycle has taken since the agent was last started
    Average cycle seconds - The average amount of time (in seconds) the evaluation cycle has taken in the last 10 cycles.
    Maximum cycle milliseconds - Same results returned as in Maximum Cycle Seconds (except in milliseconds). Provided for backwards compatibility for older agents which can't evaluate Maximum Cycle Seconds
    Average cycle milliseconds - Same results returned as in Average Cycle Seconds (except in milliseconds). Provided for backwards compatibility for older agents which can't evaluate Maximum Cycle Seconds
    Top 10 - The top 10 pieces of content the agent spends the most time on during its evaluation cycle.

Note: The Evaluation Cycle: The agent has two modes of evaluating "Action Only" and "Full Background". The agent moves into the "Action Only" mode when it needs to process an action. And, it moves back into the "Full Background" mode when it is evaluating content. The "Action Only" mode cycles are shorter in duration than the "Full Background" mode cycles. The Action Only mode cycles are counted along with the Full Background mode cycles so there is a bit of a skew (but the skew is typically negligible for a long running client with little action activity). The goal of these counters is to help point out agents that are having problems evaluating their content in an efficient manner in order prompt for further troubleshooting using client debug logging, the client usage profiler, and client diagnostics.

Duration Properties (the total amount of time, wall clock in seconds, the agent has spent evaluation since it was last restarted):
    _action duration - Total amount of time the agent has spent processing actions
    _api duration - Total amount of time the agent has spent processing api calls
    _archive duration - Total amount of time the agent has spent processing archive jobs for uploading data
    _gather duration - Total amount of time the agent has spent gathering site data
    _other duration - Total amount of time the agent spends on any other operation (this number is skewed on initial startup of agent; primarily AD information; and on Mac's application usage)
    _property duration - Total amount of time the agent has spent on processing properties.
    _quiet mode duration - Total amount of time the agent has spent in quiet mode
    _relay select duration - Total amount of time the agent has spent doing relay selection
    _relevance duration - Total amount of time the agent has spent processing relevance
    _report duration - Total amount of time the agent has spent preparing reports
    _sleep duration - Total amount of time the agent has spent in sleep mode
    _total duration - The total sum duration of time the client has spent evaluating and processing content (the sum of all previous duration property values)

Percentage Properties (percentage):
    _action percent - Percentage of time the agent has spent processing actions
    _api percent - Percentage of time the agent has spent processing api calls
    _archive percent - Percentage of time the agent has spent processing archive jobs for uploading data
    _gather percent - Percentage of time the agent has spent gathering site data
    _other percent - Percentage of time the agent spends on any other operation (this number is skewed on initial startup of agent; primarily AD information; and on Mac's application usage)
    _property percent - Percentage of time the agent has spent on processing properties.
    _quiet mode percent - Percentage of time the agent has spent in quiet mode
    _relay select percent - Percentage of time the agent has spent doing relay selection
    _relevance percent - Percentage of time the agent has spent processing relevance
    _report percent - Percentage of time the agent has spent preparing reports
    _sleep percent - Percentage of time the agent has spent in sleep mode
    _total percent - The total sum of percentages of time the client has spent evaluating and processing content (the sum of all previous duration property values). Should be close to 100%

To use the analysis, on the console menu choose File > Import, located the analysis and import it into one of your content sites. If you want to limit the number of endpoints that receive and process this analysis; first create a custom content site, subscribe only the computers you want processing this analysis and then subscribe only the endpoints you want gathering and evaluating the analysis, then import or move the analysis into this site. Once the analysis has been imported, activate it. The evaluation period for each property is set to "Every Report" by default.

Agent Performance Counters.besAgent Performance Counters.bes

Rate this page:

(0 users)Average rating

Add comments

Document information


More support for:

Tivoli Endpoint Manager

Software version:

Version Independent

Operating system(s):

Platform Independent

Reference #:

1669200

Modified date:

2014-04-02

Translate my page

Machine Translation

Content navigation