DX Application Performance Management

  • Private Community
Back to discussions
Expand all | Collapse all

what is the difference between "Number of Metrics" and "Number of Historical Metrics", and why such a big difference in values

  • 1.  what is the difference between "Number of Metrics" and "Number of Historical Metrics", and why such a big difference in values

    Posted Feb 06, 2015 08:40 AM

    We have experienced the Historical Metric Clamp a few times now.  I have cleaned some up, as well as increasing our clamp limit.

     

    But my ultimate question is why do I see a "Number of Historical Metrics" of over 1.5 million, and a "Number of Metrics" of only 192 K.

     

    If this means I have 1.3 million historical metrics I can delete, how do I figure that out?  I did a SmartStorTools "test-regex", and that seems to list listed out a little over 1 million metrics (with over 800 K being backend metrics - I can see one of our teams is somehow generating a ton of similar but slightly different SQL statements).



  • 2.  Re: what is the difference between "Number of Metrics" and "Number of Historical Metrics", and why such a big difference in values
    Best Answer

    Broadcom Employee
    Posted Feb 14, 2018 10:58 PM

    Here is the steps to change your threshold historical limit:

    1- change the limit inside "apm-events-thresholds-config" file, you can find this file under your CONFIG/ folder.
    2- here are the paremeter:
    <clamp id="introscope.enterprisemanager.metrics.historical.limit">
    <description>
    Per EM limit. Takes into account metrics with Smartstor data (i.e. live and historical metrics)
    </description>
    <threshold value="DESIRED VALUE"/>
    </clamp>
    3- Restart the collector to take effect.
    4- All collectors of your cluster must have the same value, otherwise it will brings problems.

     

    Hope it helps.



  • 3.  Re: what is the difference between "Number of Metrics" and "Number of Historical Metrics", and why such a big difference in values

    Posted Feb 23, 2018 09:49 AM

    Number of Metrics: That's the live metrics the agents are currently sending.

    Historical Metrics: That's all the metrics the Enterprise Manager has ever seen.

     

    If you start an agent with a name X, and then shut it down, and restart it with another name Y, then all X metrics are no longer live and become historical metrics, still registered in the SmartStor.

     

    Therefore, historical metrics may increase if you change the "host|process|name" often for a same agent,

    or, if you have uncleaned old stopped agents,

    or, if your applications generate metric names that change over time (typically URL or SQL metrics).

     

    Over time, the number of historical metrics naturally increases with respect to the live metrics.

     

    APM 10.5 and above handles historical metrics better with a default clamp of 30 Millions and with auto-cleaning of the Smartstor metadata.

     

    It should no longer be a problem starting with APM 10.5.