This is a multipart message in MIME format. --=_alternative 004AF0B986256E9E_= Content-Type: text/plain; charset="US-ASCII" Greetings all, I've been searching high and low for any inkling of how people are monitoring classes for long term usage data. I'm thinking something along a script to collect the data for mrtg or perhaps even an snmp module that would present them as interfaces to be polled....has anyone done any work in this area? John Dunning Assistant Director Net & Tech Svcs Wayne State College --=_alternative 004AF0B986256E9E_= Content-Type: text/html; charset="US-ASCII" <br><font size=2 face="sans-serif">Greetings all,</font> <br> <br><font size=2 face="sans-serif">I've been searching high and low for any inkling of how people are monitoring classes for long term usage data. I'm thinking something along a script to collect the data for mrtg or perhaps even an snmp module that would present them as interfaces to be polled....has anyone done any work in this area?</font> <br> <br><font size=2 face="sans-serif">John Dunning</font> <br><font size=2 face="sans-serif">Assistant Director Net & Tech Svcs</font> <br><font size=2 face="sans-serif">Wayne State College</font> --=_alternative 004AF0B986256E9E_=--
Am Monday 24 May 2004 15:39 schrieb John B Dunning/NS/WSC: > I've been searching high and low for any inkling of how people are > monitoring classes for long term usage data. For now, I'm just playing around with a patched tc, a shell script, and the RRDTool to create graphs. The rates shown by tc-statistics however don't seem to add up properly. Parent classes sometimes have higher rates than all their children together. > or perhaps even an snmp module that > would present them as interfaces to be polled I don't use snmp myself, but I think Jason Boxman (in #lartc on oftc.net) does some kind of monitoring that way. Maybe you could try asking him. ;-) Andreas
On Monday 24 May 2004 10:24, Andreas Klauer wrote: <snip> > > or perhaps even an snmp module that > > would present them as interfaces to be polled > > I don't use snmp myself, but I think Jason Boxman (in #lartc on oftc.net) > does some kind of monitoring that way. Maybe you could try asking him. ;-) Heh. Mentioned back in December on this list, I have been playing with an Net-SNMP extension[1]. Not being much of an RRDTool person, I setup a Perl script that uses Net::SNMP that is callable by Munin[2] to update an RRDTool database every five minutes. (Is that too little resolution?) [1] http://x-ray.prokon.cz/data/snmp/ [2] http://www.linpro.no/projects/munin/ I have been tracking qosBps as a GAUGE value, since deleting the root qdisc and recreating it results in qosBytes being erased. Does anyone have a preferred way of obtaining the rate information? I imagine it should be converted to bits? Any other hints for producing useful graphs with RRDTool for traffic shaping statistics? (Mine are rather bland... Bulk is full 95% of the time and there are only four classes, all sampled as GAUGE at a five minute resolution...) > Andreas -- Jason Boxman Perl Programmer / *NIX Systems Administrator Shimberg Center for Affordable Housing | University of Florida http://edseek.com/ - Linux and FOSS stuff
On Monday 24 May 2004 19:01, Jason Boxman wrote: > On Monday 24 May 2004 10:24, Andreas Klauer wrote: > <snip> > > > > or perhaps even an snmp module that > > > would present them as interfaces to be polled > > > > I don't use snmp myself, but I think Jason Boxman (in #lartc on oftc.ne= t) > > does some kind of monitoring that way. Maybe you could try asking him. > > ;-) > > Heh. > > Mentioned back in December on this list, I have been playing with an > Net-SNMP extension[1]. Not being much of an RRDTool person, I setup a Pe= rl > script that uses Net::SNMP that is callable by Munin[2] to update an > RRDTool database every five minutes. (Is that too little resolution?) > > [1] http://x-ray.prokon.cz/data/snmp/ > [2] http://www.linpro.no/projects/munin/ > > > I have been tracking qosBps as a GAUGE value, since deleting the root qdi= sc > and recreating it results in qosBytes being erased. Does anyone have a > preferred way of obtaining the rate information? =20 Trust rrd, just update the counters and rrd calculates the rate. You loose= =20 some values if you reset the tc stats, but will not happen that much. > I imagine it should be=20 > converted to bits? Any other hints for producing useful graphs with > RRDTool for traffic shaping statistics? You can stack the rates of the child classes so the sum is the parent rate.= =20 You can also stack the rates of the child classes in % so the sum is 100. = I=20 can help you with this because I did it this before. Stef =2D-=20 stef.coene@docum.org =A0"Using Linux as bandwidth manager" =A0 =A0 =A0http://www.docum.org/
On Monday 24 May 2004 16:24, Andreas Klauer wrote: > Am Monday 24 May 2004 15:39 schrieb John B Dunning/NS/WSC: > > I've been searching high and low for any inkling of how people are > > monitoring classes for long term usage data. > > For now, I'm just playing around with a patched tc, a shell script, and > the RRDTool to create graphs. The rates shown by tc-statistics however > don't seem to add up properly. Parent classes sometimes have higher rates > than all their children together. > > > or perhaps even an snmp module that > > would present them as interfaces to be polled > > I don't use snmp myself, but I think Jason Boxman (in #lartc on oftc.net) > does some kind of monitoring that way. Maybe you could try asking him. ;-) There is a patch for snmp so you can get tc stats with snmp. I created som= e=20 scrpits to graph this information with rrd. See docum.org. Stef =2D-=20 stef.coene@docum.org =A0"Using Linux as bandwidth manager" =A0 =A0 =A0http://www.docum.org/
Hi, -> There is a patch for snmp so you can get tc stats with snmp. I -> created some -> scrpits to graph this information with rrd. See docum.org. I've patched net-snmpd with patch that I see on docum.org: http://x-ray.prokon.cz/data/snmp/ # ./configure --with-perl-modules --with-mib-modules=qos # make # make install but when I try to exectue "proces.pl" it send me: "ERROR: Message size exceeded buffer maxMsgSize" How ever, when I try to test QoS with snmpget, like in README file indicates: # snmpget -c macwww localhost enterprises.18756 snmpget: No securityName specified # snmpget -c macwww localhost -u root enterprises.18756 snmpget: Unknown user name Is a snmp patch problem related or "proces.pl" related ? Thank you !!! Andres