[mrtg] FAQ

[alex at ergens.op.Het.Net]

Frequently answered questions on the MRTG list.

Request to regular list members:  please help me out to complete this FAQ.  

This is a small FAQ.  I really want to keep it small so that it can
be posted on the list frequently.  If you don't want to be bothered,
just ignore the list for subjects starting with FAQ.
Comments are very welcome.  mailto:alex at ergens.op.Het.Net


Current content:
Q: What is this rrdtool people are talking about?
Q: MRTG seems to insist of always producing 2 graphs.
   Is it possible to get only one graph?
Q: What is the meaning of the numbers in the log file?
Q: I didn't understand when to use gauge/absolute?
Q: My statistics seem to display way too low values?
Q: How can I monitor <whatever> ?


Q: What is this rrdtool people are talking about?
A: RRDtool can be seen as the successor for "rateup" (which is part
   of the MRTG tool).  It can do more and uses lesser resources.
   See http://ee-staff.ethz.ch/~oetiker/webtools/rrdtool/
   Also look into "14all" which can be found in the contrib dir.

Q: MRTG seems to insist of always producing 2 graphs.
   Is it possible to get only one graph?
A: Yes.  However, you still need to process two OIDs.  If you monitor
   the same value twice, you should get identical results.  This way,
   you get a green block with a blue ceiling.  You would also like to
   produce one line of text below the graph, this is described in the
   documentation.  Look for "Note, if LegendI or LegendO are" ...
   ***See also the next question.***
Exception:
   Sometimes you won't get the same result back if you query the same
   OID twice.  NT processor load seems to be one of them.  Perhaps you
   can find a value that will always be zero; use that as your second
   OID.  Or, write a script that gets and preprocesses the value for
   you so that it will be returned twice.

Q: What is the meaning of the numbers in the log file?
A: This question indicates that you didn't read the documentation available
   in the "doc" directory.  Either you didn't read "README" at all or you
   ignored the lines saying 
      "In any case, make sure you read the mrtg documentation
       in the doc subdirectory. Especially the manual.txt."
   This is a serious error condition and there does not seem to be an
   easy solution to it.
Exception:
   Sometimes this error occurs when a question is poorly asked.  If you
   ment to ask "I didn't understand it when I read it" then do so.  This
   is, by the way, valid for any part of the documentation.
   Don't be afraid to ask; as long as you *did* read the documentation.
   If you did, and if we know that, we're glad to help.
   If you ask a question for something that is described in the
   documentation, and don't make it clear that you actually did read it,
   get ready to be flamed.  Your fault and your loss.

Q: I didn't understand when to use gauge/absolute?
A: Normally, MRTG would expect growing counters.  When it reads them,
   it has to subtract the previously measured value from it to calculate
   the delta (the difference).  Example: the first time it reads 1000
   from the counter, five minutes later it reads 1600.  This is an
   increase of 600 in five minutes.  This is a normal counter and you
   do not need absolute nor gauge.  MRTG uses the delta (600) and the
   delta time (in this case 300 seconds) and calculates 2 per second as
   the current rate.
   Some counters are actually reset when you read them.  In our example,
   the second sample would not be 1600 but rather 600.  This means that
   MRTG does not need to subtract the values and we tell that to MRTG
   by using the option "absolute".  The rate is still dependent on the
   time lapsed (300 seconds) and the rate will be 2 per second again.
   Some values we measure are not really counters.  They are not depending
   on the number of seconds since the last time we measured.  Think of
   temperature.  If MRTG reads a value, it does not need to subtract a
   previous value and it also need not to divide by time.  This is when
   you need to use the option "gauge".

Q: My statistics seem to display way too low values?
A: The are a couple of things to check.  First of all, make sure your
   device does return the proper values to MRTG.  You can do so by
   using tools such as "snmpget".  How to use them is not something for
   *this* FAQ.
   Secondly, be aware of the problem with 32-bit counters which MRTG is
   using.  32 bit will allow for a maximum value of 4294967295 and when
   divided by 300 (seconds) this is about 14 megabytes or 114 megabits.
   If your interface transports data near or above this rate, look for
   another tool.  Look into rrdtool and one of its available front ends.
   A temporary work around may be to query more often.  Just call the
   script every minute (no need for *any* modifications to the config
   file) and the problem is near 4294967295/60 instead of 4294967295/300
   Beware!  This may not be possible on your platform, make sure it can
   handle the load!  You have been warned.  In any case if you need this,
   start using RRDtool ASAP.

Q: How can I monitor <whatever> ?
A: Before you ask this to the MRTG people or process, make sure you can
   collect the data.  You can do this with or without SNMP and it is not
   in the scope of the MRTG list to provide these resources.  Sometimes
   these kind of questions are asked and even answered, but they are not
   MRTG specific.  I'll give one example:  suppose you want to monitor
   the cpu usage of a Cisco router.  Find out how to query this value and
   you should come up with the OID "1.3.6.1.4.1.9.2.1.58".  When you use
   this OID, you also need to specify an instance (0 in this case) and you
   need to use two OIDs (see elsewhere in the FAQ).  Check your OID and
   router config by using, for instance, snmpget:
   snmpget router public .1.3.6.1.4.1.9.2.1.58.0
   If you won't get an answer, no need to blame it on MRTG as you didn't
   even use it yet.  When you do get an (correct!) answer, you can build
   a target into your config file:
   Target[x.cpu]: 1.3.6.1.4.1.9.2.1.58.0&1.3.6.1.4.1.9.2.1.58.0:public at x
   Options[x.cpu]: gauge
   Add other options etcetera as needed.
   If this didn't make any sense to you at all, learn how to use SNMP
   and return here when you have.  There are plenty places on the net
   where you can find out what SNMP is and how it works.

--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg