Torrus

From Wikitech
Jump to: navigation, search

Torrus is an SNMP monitoring/graphing system. It's like a much more advanced version of Cricket.

Configuration and setup

It's installed on netmon1001. The configuration directory is /etc/torrus/, with the configuration of data sources and graphing configuration in XML files under /etc/torrus/xmlconfig/. Some global Torrus options are set in /etc/torrus/conf/torrus-siteconfig.pl. Torrus display style has been modified via files in /etc/torrus/templates; these files are needed for it to work as they are referenced from the data sources XML files.

Discovery

A daily cron job is run to update the list of ports on switches and routers, using the files in /etc/torrus/discovery/. The cron job is in /etc/cron.daily and looks like this:

#!/bin/sh
DDXFILES="coreswitches.ddx accessswitches.ddx"
TREES="Network"

for ddx in $DDXFILES
do
	torrus devdiscover --in=$ddx
done

for tree in $TREES
do
	torrus compile --tree=$tree
done

Common problems

Deadlock problem

When Torrus stops working in any way, that's usually because it ends up deadlocked. The web processes and the collector processes work on the same DBM files in /var/lib/torrus/db. Sometimes they get into a deadlock, presumably because some process died and didn't reduce a lock. Here's the steps to resolve the issue:

service torrus-common stop
db5.1_recover -h /var/lib/torrus/db
sudo -u Debian-torrus torrus compilexml --all --verbose
service torrus-common start

More detailed information on how to resolve this can be found here.