[mrtg] Re: provisioning question for a new MRTG server

[Jerry Heidtke]

I'm using mrtg running under Windows NT 4.0, on an old Compaq Proliant 2500
with dual 200 MHz Pentium Pro cpus. The box has 448 MB of RAM. I have about
5,000 active rrd files. Almost all of these are updated every five minutes;
about 100 are updated every minute.

The config files are split across 8 instances of mrtg, each launched through
cron (I use the cron.exe from the aintx toolset).

Each 5 minute instance of mrtg updates between 500 and 1000 rrd files. These
take from 30 to 90 seconds to run, as long as all the devices are
responding. When a device is offline, it can take nearly the entire five
minute cycle to complete a run. This doesn't impact cpu or memory
utilization much, since mrtg is just waiting for responses. I've adjusted
the snmp timeouts to minimize the impact of this.

CPU utilization averages about 35-40%. Graph generation (using Apache and
14all.cgi running under mod_perl) is slower than I'd like, but usable. Don't
even think about doing this with IIS without expensive add-ons. It takes
about 5 seconds to deliver a page with 12 graphs on it. There are ususally
2-3 people looking at graphs pretty regularly, with another 5-6 occasional
users. If there were many more people looking at graphs, this box would fall
over pretty quickly. With low usage, I could probably squeeze 50% more data
collection onto it.

Disk i/o doesn't seem to be a problem here, the box has a older RAID 5 array
with 3-9GB fast-wide (not ultra) SCSI drives (I think 7200 rpm), the array
controller has 4 MB of cache, and the controller CPU utilization is under
30% handling over 100 operations per second. Latency on the controller
(average length of time commands wait because the controller is busy) is
about 0.2 seconds. I wouldn't want to try this with ATA/IDE drives.

Network utilization is way less than 1% of a fast-ethernet connection.

I will be upgrading the box soon to a new server, probably a P3 1.2 GHz, 1
GB RAM, ultra-wide SCSI array with 10 or 15,000 rpm drives. I would then
split the config files across more mrtg instances so that devices that are
down don't interfere as much with collecting data from everything else.
Graph generation should be much quicker. This kind of box would probably
handle your expected load without breaking a sweat, even running Windows.

Most of the time mrtg is running, it is waiting for responses to snmp
queries. The key to having a large installation of mrtg seems is to get
several instances running in parallel (probably using forks on 'nix), and to
stagger the work over the collection period using different schedules for
different parts of the overall task. Also, adjust snmp timeouts and retries
to handle non-responsive devices.

Eventually, I want to move the whole mess to cricket, since manually
maintaining several hundred mrtg config files is a major pain. That will be
interesting...

Hope this helps.

Jerry

-----Original Message-----
From: Diana Eichert [mailto:deeiche at freeshell.org]
Sent: Wednesday, June 05, 2002 2:59 PM
To: MRTG-list (E-post)
Subject: [mrtg] provisioning question for a new MRTG server



I've been running several MRTG servers against several hundred Cisco network
devices 
for a number of years, monitoring selected ports on Catalyst due to the
sheer 
number of ports on the Cats.  We have almost eliminated all hubs in our
network 
environment.

Management has recently gotten behind the use of modeling tools to plan for 
future network upgrades.  One of these tools can use real MRTG data to build
the 
model.  They'd like to monitor all the Catalyst ports, over 10,000 right
now, so 
when an enduser group comes in requesting a faster connection we will have
longterm 
historical data to assist in planning.  Now we're looking at monitoring a
much 
greater number of ports.

I've looked for documentation on sizing MRTG w/RRDTOOL on the web but
haven't come 
up with anything, so far.  I believe that the largest limitation is disk IO
because 
you have to read in, then write out each rrd file.

Opinions based on real experience is appreciated.  Pointers to URLs where
this has 
discussed before is welcome also.

thanks

diana eichert

--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi


Confidentiality Notice: This e-mail message, including any attachments, is 
for the sole use of the intended recipient(s) and may contain confidential 
and privileged information.  Any unauthorized review, use, disclosure or 
distribution is prohibited.  If you are not the intended recipient, please 
contact the sender by reply e-mail and destroy all copies of the original 
message.


--
Unsubscribe mailto:mrtg-request at list.ee.ethz.ch?subject=unsubscribe
Archive     http://www.ee.ethz.ch/~slist/mrtg
FAQ         http://faq.mrtg.org    Homepage     http://www.mrtg.org
WebAdmin    http://www.ee.ethz.ch/~slist/lsg2.cgi