DX NetOps

  • 1.  replacing/deleting outliers

    Posted Feb 05, 2015 07:36 AM

    Anyone have a method for replacing or deleting outliers? For example, tracking tens of interfaces over a month. One interface reports utilization of 15 billion %. Makes the chart unusable, plus there's no sense in reporting such a value. I'd like to delete the value or replace with an appropriate value.

  • 2.  Re: replacing/deleting outliers
    Best Answer

    Posted Feb 05, 2015 09:32 AM

    I can't help you with deleting the existing data, but this will most likely help going forward:


    On each Data Collector in the directory $IMROOT/apache-karaf-2.3.0/etc, create a new file named "com.ca.im.dm.snmp.collector.SnmpCollector.cfg", and insert this single line:




    The parameter is picked up automatically, so no restart is required.



    By default the Data Collector tries to be smart about calculating counter roll-overs.  This is a great idea and good intentions on CA's part -- however, SNMP agents have the potential to be buggy and we came across many vendors/platforms that like to roll-over at random times.  The Data Collectors happily calculate the amount of data that would've been needed to account for a legitmate roll-over, and thus you're left with 15,000,000,000%.


    We've been in really good shape since applying this parameter, as we would rather see a gap in data than bogus data.

  • 3.  Re: replacing/deleting outliers

    Posted Feb 09, 2015 01:22 PM

    Thanks. I'll give it a shot.

  • 4.  Re: replacing/deleting outliers

    Posted Feb 09, 2015 04:06 PM

    Thanks Justin.  That's a nice little hack.  It would be really cool if this file were to be included with the default install but have that line commented out.  That would make it a lot easier for folks to change the setting if desired.

  • 5.  Re: replacing/deleting outliers

    Posted Mar 12, 2015 02:04 PM

    The answer provided helps to keep outliers from getting created moving forward but there is still the question of how to get remove or mitigate an outlier that has already been processed.  Is there a method to mitigate existing outliers?