gtuhl: startup IT and software development

I recently installed Zenoss Core (available at zenoss.com) at the office for the purpose of monitoring the handful of machines we have there. My goal with this post was to be exhaustive but I decided instead to give a quick summary and some tips as the documentation is quite good already. A community site with forums, a blog, and a wiki can be found at community.zenoss.com and it is pretty lively. I’ve asked 2 questions thus far and received prompt and helpful responses and discussion in return.

In the past my tools for monitoring machines have consisted of Nagios and patchworks of shell scripts. Shell scripts aren’t particularly manageable for a large number of machines and I found Nagios to be too difficult to configure and learn. Nagios is a really powerful tool used by a lot of people though so don’t discount it based solely on my unqualified opinion. To learn more about Nagios go here.

That said, I was trying to find a new approach and slashdot recently ran a post about Zenoss. After giving it a shot and finally getting it installed I have to say I really love this thing for monitoring network devices and servers. It is fairly easy to install and learn, has a great interface and charting out of the box, is incredibly configurable and customizable, has an active and helpful community, great documentation, and if needed can run Nagios plugins.

Installing on OpenSuse 10.2 64bit

As of a few days ago 64bit rpms are not available for installing Zenoss so you will need to go with the source tarball found on the download page here. Once you have the source where you want it on the machine that will do your monitoring you can use the instructions here to perform the installation. There really isn’t anything missing from their instructions, but I didn’t take everything literally and it made things take longer. If you have trouble with the install be sure to check the zenbuild.log file. Here are the issues I faced.

• If during the install you get messages along the lines of wrong ELF class: ELFCLASS32 then you are probably trying to use the 32bit rpm to install Zenoss on a 64bit OS. Go grab the source tarball instead.
• You need MySQL 5.0.22 or greater. Though this post doesn’t cover the rpm installation in depth if you are able to use the rpm and it complains about this dependency even when you are 100% certain you have a qualified MySQL version installed (such as happened for me on a 32bit OpenSuse box) just force it with --nodeps and you should be fine.
• You really do need Python 2.3.6 or 2.4. If you use Yast with OpenSuse you will not be able to use the Python it provides (its 2.5). If this is the case go to the 2.3.6 download page on python.org, grab the tarball and do a ./configure, make, make install. You should now be set. Make sure running which python points you to the 2.3.6 version. You should not need to uninstall Python 2.5 from your machine. When I tried to do this Yast complained about all sorts of dependencies.
• If during the install you get messages related to mkzopeinstance.py then you have Python 2.5 installed and not 2.3.6 or 2.4. You need to go grab and install the correct version.
• If during the install you get messages containing undefined symbol: Py_InitModule4 it probably means the Zenoss installer is seeing and running a Python 2.5 interpreter and trying to load Python 2.3.6 or 2.4 modules. Make sure any symlinks or references to Python 2.5 are converted.

Other Tips

• If you get stuck during the install feel free to post here or better yet ask in the community forums for Zenoss.
• Once everything is up and running you will want to work through the quick start guide on the Zenoss documentation page - obviously skipping the vmware related stuff if you’ve just installed your own copy.
• If when adding hosts to monitor Zenoss gives you a no snmp found for ip = x.x.x.x you need to make sure an snmp agent is installed and accessible on the host. I found these instructions to be quite helpful to make it work quickly. Skip down to step 19 and remember those instructions are for Ubuntu so you’ll probably be using yast on OpenSuse instead of apt-get.
• One thing that threw me briefly with the snmpconf tool was the snmpd.conf file it generates is written to whatever directory you ran the tool from. Make sure you cd to /etc/snmp/ before running the tool or copy the generated .conf file to its correct place.
• After changing your snmp configuration you will want to do a /sbin/service snmpd restart.
• To monitor windows machines through snmp you will need to enable it as described here.

The only other tip I can offer is to think about the monitoring as being done from inside of the device or machine. Through snmp Zenoss will be able to check for actual processes, disk usage, and machine information. My experience with monitoring applications like this was so limited that I had the mindset of wanting to monitor everything through IP ports. This isn’t the way to think about it - though Zenoss allows this as well. By setting up some processes and alerting rules you can have some pretty effective monitoring in place very quickly and you will have more options doing it this way.

No matter how small your office or how limited your production environments your customers (whether internal or external) should never be able to surprise you with news of down time. Zenoss is a smart, effective tool that can be used to monitor 1 or tons of devices and I have been very pleased with it thus far.