Observium is a great tool for network monitoring, reporting and analysis. It’s community-supported version is free and very comprehensive, giving administrators deep visibility into their network links and devices including device-level, throughput, port information and more! I’ve successfully implemented it for switches, network devices, servers and storage arrays including NetApp products and been very happy with the results.

While the install guide at http://www.observium.org/wiki/RHEL_Installation is pretty comprehensive, there are a few gotchas that I’ve seen in a few installs that needed repair.

For one, graphs not populating- this is generally admin error when loading cron jobs. On the instructions, they advise to not use “crontab -e” to insert the jobs… This can easily be modified by cleaning up the syntax and removing the user command:

33 */6 * * * /opt/observium/discovery.php -h all >> /dev/null 2>&1
*/5 * * * * /opt/observium/discovery.php -h new >> /dev/null 2>&1
*/5 * * * * /opt/observium/poller-wrapper.py 1 >> /dev/null 2>&1

Once modified, modify crontab as the root user or by executing “sudo crontab -e -u root” as a user with sudo privileges.

Poller issues are also common, given that the default install uses only one by default. On most systems, this isn’t optimal and there is significant resources to allow more instances. To run more instances in parallel, modify the crontab entry for the poller, replacing the bold number to suit your environment:

*/5 * * * * /opt/observium/poller-wrapper.py 20 >> /dev/null 2>&1

If this causes IO issues, consider tuning it to your environment or using a RAM disk or faster storage.

MySQL tuning is also essential for larger environments, consider moving MySQL to a dedicated instance within the same LAN if possible or to a separate volume, again with faster storage such as an SSD, SAS drives and/or optimized RAID such as RAID 10 or RAID 6 depending on your system.

Do you have any tips or tricks to share? Feel free to comment with them! If you haven’t given Observium a try, visit the website at http://www.observium.org for a live demo, documentation and downloads.

Working on moving rrdtool graphs over to Cacti after enough fun with other platforms… It was a bit of an adventure!

Using CentOS 6.2 x86, I installed the prereqs and Cacti package as instructed. At first, there were no graphs. After a bit of playing around, set a log on the cron entry for poller.php – not much info aside from a single entry stating that the user was invalid.

I tested the script as the Cacti user and guess what- ended up with rrd files in the rra folder, including graphs within Cacti! Deleted all references to the cronjob under root (including the one in /etc/cron.d/) and recreated it under my Cacti user. Deleted the definitions within Cacti and deleted the rrd files in the rra folder. What do you know… graphs again, via cron!

This led to another problem- no data. -nan shown for all data. After 30 minutes of searching the web, found a helpful suggestion that fixed the issue! After install, run a “chmod +s” on the poller script- suddenly, data!

What fun… now to play with the system!