Re: [Devmon] if_err stats

[Mario V. <mar...@gm...>]

I'm working on standard, unmodified templates.  In particular with 3750,
4948 and 2801 devices.


Thanks for the advice on the 'multigraphs=disk,if_load,if_err,temp' to
CGI_SVC_OPTS in hobbitcgi.cfg or
equivalent'.  It's already a step forward.


Regarding the thresholds;  in my case, rather than wanting to check error
packets as a percentage of bandwidth, I want to be alerted if and when an
error is detected at all.  ¨Perhaps a threshold check with the  "*
{ifInErrors}"* and "*{ifOutErrors}" *will give me what I need.
I'll play around with it.


Thanks again,
Mario



On 25 January 2012 15:15, Buchan Milne <bg...@st...> wrote:

> On Tuesday, 24 January 2012 13:26:01 Mario Valetti wrote:
> > Hi,
> >
> > When selecting 'network device | <if_err>' column, the resultant table
> > shows total interface errors since the device start / reset.
>
> Can you provide the name of the template you are using for the device in
> question (if it is a standard, unmodified template), or the template itself
> you are using (if not standard unmodified)?
>
> > I'm
> > only interested in checking if errors occured within the previous 24 or
> > 48hrs (for example).
>
> The transforms file on most of the if_err tests includes:
>
> # Get bit speed delta (so we dont have to provide custom delta limit)
> ifInOps         : DELTA         : {ifInOctets}
> ifOutOps        : DELTA         : {ifOutOctets}
> # Convert our octets delta into bits per second
> ifInBps         : MATH          : {ifInOps} x 8
> ifOutBps        : MATH          : {ifOutOps} x 8
> # Do delta transform on all error counters
> ifInEps         : DELTA         : {ifInErrors}
> ifOutEps        : DELTA         : {ifOutErrors}
> # Perform error to traffic percentage calculations
> ifInErrPct      : MATH          : ({ifInEps} / {ifInBps}) x 100
> ifOutErrPct     : MATH          : ({ifOutEps} / {ifOutBps}) x 100
> # Create an alias in a bracketed box, or nothing if alias is blank
> ifAliasBox      : REGSUB        : {ifAlias} /(\S+.*)/ [$1]/
>
> And the thresholds:
> # Create thresholds for all the error rate counters
> # oid name      : color         : limit : Error message
>
> ifInErrPct      : yellow        : 5     : {ifName}{ifAliasBox} - High input
> error rate ({ifInErrPct}%)
> ifInErrPct      : red           : 10    : {ifName}{ifAliasBox} - Very high
> input error rate ({ifInErrPct}%)
> ifOutErrPct     : yellow        : 5     : {ifName}{ifAliasBox} - High
> output
> error rate ({ifOutErrPct}%)
> ifOutErrPct     : red           : 10    : {ifName}{ifAliasBox} - Very high
> output error rate ({ifOutErrPct}%)
>
>
> So, all alarming is for Error packets per second / Bits per second (so,
> unfortunately, not a real percentage) averaged over the poll period.
>
> It may instead be better to make it 'Error packets per second' / ('Unicast
> pps' + 'Multicast pps' + 'Broadcast pps') to give a real percentage.
>
> However, your assertion that the table holds the total errors since counter
> reset is false (at least for most if_err tests).
>
> Feel free to send a patch ...
>
> > I know the if_err graphs are available under the "trends" column, but is
> it
> > possible to view this period for errors directly in the table within the
> > if_err column check?
>
> You can get the graphs on the if_err page as well by adding e.g. '--
> multigraphs=disk,if_load,if_err,temp' to CGI_SVC_OPTS in hobbitcgi.cfg or
> equivalent.
>
>
> Regards,
> Buchan
>