[mrtg] Graphing Uptime

Fri Sep 28 22:37:16 CEST 2018

No, SNMP returns a pure integer. The BER module used by MRTG contains the code that formats the integer into a string that is almost identical to what the snmpwalk command produces. Look for this line in the MRTG script:

#LH        local $BER::pretty_print_timeticks = 0;

(line 2307 at the moment on Github)

If you move this to the start of sub getsnmparg[1] and liven it (and remove local), the HTML page's uptime reverts to a pure integer -- the code is deliberately choosing to render the integer as a string prematurely, HOWEVER, the graph still fails to show any data, so there must be some other reason -- I need to switch logging back on and see what's wrong with it this time! (Also, just enabling that line where it stands doesn't have any effect -- probably because it only seems to work if you fetch one value at a time and uptime is always fetched in pairs, either twice because of the whole "noo" malarkey, or in a pair alongside sysName.)

Formatting of gauge and what not (for which I don't recall ever reading a meaningful explanation) is an irrelevance until the data stops being mangled or mishandled. Once the program can actually recognise the integer it's being given, I can fix the graph if I have any pressing need.

Calling of BER's pretty_print_timeticks should be done only where required -- but understanding the way BER is involved and why there is premature formatting would take ages.

[1] I declare all my "use" statements at the top, but MRTG's code makes it hard to see where BER is initialised, to make sure that this is called after something else sets it to true in the first place, so I left the call in the function where it's relevant.

On 2018-09-28 21:21, Edwin A. Epstein III wrote:
> Hi Daniel,
> 
> Yes that example was horribly bodged. I haven't inspected the code, but I suspect MRTG works with the value returned by SNMP. For example, I receive this:
> 
> SNMPv2-MIB::sysUpTime.0 = Timeticks: (105630500) 12 days, 5:25:05.00
> 
> That cannot be graphed because it is not a number. Everything that MRTG graphs must be turned into some number. The example is also horribly bodged because it's trying to use a bandwidth graph instead of a gauge. MRTG provides for graphing values like CPU load, Memory, and Free disk space. You really want to grab the most recent book as it will tell you how to construct these. I'll give you an example:
> 
> Target[the_graph]:        1.3.6.1.4.1.32050.2.1.27.5.1&1.3.6.1.4.1.32050.2.1.27.5.1:snmp_community_name at 10.0.0.1:::::2 * -1.1034882
> Options[the_graph]:       unknaszero,gauge,growright,nopercent,expscale,noo
> SetEnv[the_graph]:        MRTG_INT_IP="No Ip" MRTG_INT_DESCR="n/a"
> Colours[the_graph]:       ORANGE#dd8811,NONE#000000,VIOLET#0000ff,DARK GREEN#006600
> Title[the_graph]:         Voltage Monitor
> MaxBytes[the_graph]:      850
> AbsMax[the_graph]:        850
> XSize[the_graph]:         600
> 
> All of these directives are explained in the book. The two most important ones are the Target and Options directives. The gauge option is what makes it a gauge graph, and the noo option suppresses one side of the graph (input or output). With the directives you can construct your own custom graph with correct units for uptime, and a scale that will make sense. You can control titles, legend values, etc.  
> 
> Your first issue is how to convert 'Timeticks: (105630500) 12 days, 5:25:05.00' to a number. I would suggest graphing the hours of uptime. Even after a few years of uptime, the value itself will be less than 100,000 and probably graph well over time. 
> 
> MRTG provides for pre-processing of SNMP values before they are passed to MRTG. I'm performing math before I use the voltage value. Since I'm pretty sure that the math is any valid perl statement, you might be able to get away with Perl. That being said, you may be best served by simply creating your own data collection plug-in, which is thankfully easier done than said. Straight from the book:
> 
> Target[ezwf]: `/usr/local/bin/mrtg-scripts -a 1`
> 
> All you need to is create a bash script that pipes your snmpwalk output into a awk, and then convert the returned value into the number of hours. Timeticks can be converted to hours: Hours = Timeticks / 100 / 60 / 60. 
> 
> Something like:
> 
> uptime=`snmpwalk -v1 -c public 10.0.0.1 SysUptime | awk -F'[()]' '{print $2}'`
> let hours=uptime
> let hours=$hours/100/60/60
> echo $hours
> 
> If you run that bash and pass it into MRTG, with directives to create a gauge type graph, you should get a fairly nifty uptime graph. With correct units for time as a bonus. 
> 
> 
> 
> Sincerely, 
> 
> Edwin A Epstein, III 
> Rhinobee Internet Services 
> 707.237.7504 ext 209 
> 707.737.0288 Mobile
> 
> ----- Original Message -----
> From: "mrtg-request" <mrtg-request at lists.oetiker.ch>
> To: "mrtg" <mrtg at lists.oetiker.ch>
> Sent: Friday, September 28, 2018 3:00:02 AM
> Subject: mrtg Digest, Vol 132, Issue 1
> 
> Send mrtg mailing list submissions to
> 	mrtg at lists.oetiker.ch
> 
> To subscribe or unsubscribe via the World Wide Web, visit
> 	https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
> or, via email, send a message with subject or body 'help' to
> 	mrtg-request at lists.oetiker.ch
> 
> You can reach the person managing the list at
> 	mrtg-owner at lists.oetiker.ch
> 
> When replying, please edit your Subject line so it is more specific
> than "Re: Contents of mrtg digest..."
> 
> 
> Today's Topics:
> 
>    1. Graphing uptime (Daniel Beardsmore)
> 
> 
> ----------------------------------------------------------------------
> 
> Message: 1
> Date: Fri, 28 Sep 2018 10:06:23 +0100
> From: "Daniel Beardsmore" <resident at telcontar.net>
> To: <mrtg at lists.oetiker.ch>
> Subject: [mrtg] Graphing uptime
> Message-ID: <041801d4570a$88f67650$9ae362f0$@telcontar.net>
> Content-Type: text/plain; charset="us-ascii"
> 
> Hello
> 
>  
> 
> I can see that graphing uptime is possible, as you can see here:
> 
>  
> 
> http://www.hotelsvillegia.com/mrtg/uptime.html
> 
>  
> 
> The HTML pages report uptime in the format: "163 days, 21:07:10"
> 
>  
> 
> If I check manually, I get this:
> 
>  
> 
> snmpget -v2c -c somecommunity somehost 1.3.6.1.2.1.1.3.0
> 
> DISMAN-EVENT-MIB::sysUpTimeInstance = Timeticks: (1415941565) 163 days,
> 21:10:15.65
> 
>  
> 
> The format is almost the same, but the latter contains the full centisecond
> accuracy. You do nonetheless get the raw number included.
> 
>  
> 
> Now, using this in MRTG yields:
> 
>  
> 
> 2018-09-27 19:56:04 -- 2018-09-27 19:51:33: WARNING: Expected a number but
> got '163 days, 7:17:10'
> 
>  
> 
> Looking at the source code, I cannot determine quite how uptime is
> processed. It seems odd that the format is almost the same (without the
> centiseconds), which suggests (along with other code) that MRTG receives
> pre-formatted output, and then has to scrape out the useful bits. (Which is
> just plain horrible if this is true.)
> 
>  
> 
> Am I right in thinking that MRTG presently has no way to extract the raw
> figure here? It seems that the SNMP library is formatting the data
> prematurely and MRTG just works with that preformatted value as it suits its
> own purposes, but that you cannot get the raw data out if you choose, for
> example if you want to record uptime as a graph for checking for reboots.
> 
>  
> 
> In the example posted, I suspect that was bodged to get that to work.
> 
>  
> 
> Regards
> 
>  
> 
> Daniel. 
> 
> -------------- next part --------------
> An HTML attachment was scrubbed...
> URL: <http://lists.oetiker.ch/pipermail/mrtg/attachments/20180928/8eacb971/attachment-0001.html>
> 
> ------------------------------
> 
> Subject: Digest Footer
> 
> _______________________________________________
> mrtg mailing list
> mrtg at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
> 
> 
> ------------------------------
> 
> End of mrtg Digest, Vol 132, Issue 1
> ************************************
> (null)
> 
> _______________________________________________
> mrtg mailing list
> mrtg at lists.oetiker.ch
> https://lists.oetiker.ch/cgi-bin/listinfo/mrtg
> 