[mrtg-developers] Re: Blocking Outages

[Garry W. Cook]

I have also experienced this problem, since switching to a single mrtg
daemon process for all 75 of my routers. Originally I forked the process
into 12 instances, and after experiencing the problem, I have tried forking
into 32 instances and the problem still occured.

Previously, I had the the 75 routers split into 5 different regions, with a
seperate process forking into 4 instances for each region. If two or more
routers went down in a single region, the entire mrtg process for that
region would usually die, causing me to lose all data. Although the process
would survive if I only lost one router.

In both instances, I have a cron job that runs every five minutes checking
for the mrtg process(es), and restarting if necessary. However, the process
will start and then die again when it reaches the outage.

Today I tried your suggestion about modifying the SNMP retry attempts,
although I could not get the SnmpOptions global parameter to work correctly.
Has anyone had success with this?

I have just increased the number of forks to 48. We'll see what happens the
next time there is a router down.

Garry W. Cook
garry at cookbros.net
303.359.8322

> -----Original Message-----
> From: mrtg-developers-bounce at list.ee.ethz.ch
> [mailto:mrtg-developers-bounce at list.ee.ethz.ch]On Behalf Of Brian E.
> Seppanen
> Sent: Thursday, May 24, 2001 6:39 AM
> To: mrtg at list.ee.ethz.ch
> Cc: mrtg-developers at list.ee.ethz.ch
> Subject: [mrtg-developers] Blocking Outages
>
>
>
> I think I've asked this question before but I don't think anyone has ever
> indicated any possible solutions.
>
> I currently graph 2200 targets and that number will grow.  We're using
> MRTG 2.9.11 on a DELL Poweredge 2300 Server with Dual PIII 500 and 512 MB
> of RAM.  We're running MRTG in daemon mode with the possible forks set to
> 4.
>
> As it is currently if one of our monitored routers goes down for any
> length of time it blocks all other graphs from graphing.  I've tried
> modifying the default SNMP session values to only doing a single retry, to
> hopefully ease the blocking.  So far I haven't noticed a benefit in terms
> of blocking and graphing outages.
>
> I have been told that we need to monitor on each router in/out
> traffic,collisions,crc's, and some signal/noise on some routers.  Perhaps
> this is trying to do to much on one router, however, I need to see if it's
> possible to set up mrtg so that it graphs seemlessly when one router goes
> down.
>
> Has anyone seen anything similar to this?  Has anyone resolved this?  Is
> there any way of integrating target configs into mysql or any sql database
> such that a wrapper program could determine if there is a non-responding
> device and take it out of rotation, while periodically performing a check
> to find out if the device is avaliable again?  I wish I could write it
> myself.
>
> Can perl take advantage of multiple threads?
>
>
> Brian Seppanen
> seppy at chartermi.net
>
>
> --
> Unsubscribe
mailto:mrtg-developers-request at list.ee.ethz.ch?subject=unsubscribe
Help        mailto:mrtg-developers-request at list.ee.ethz.ch?subject=help
Archive     http://www.ee.ethz.ch/~slist/mrtg-developers



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