There is a tendency sometimes to think that system monitoring tools like munin, zenoss, zabbix or nagios are mainly for the sysadmin. We think the people who can benefit from monitoring is considerably bigger.
Support
While it may be true that tech/software support sit close to the sysadmins, or in smaller companies might even be the same people, the job is a different one. In the commercial software world great support is a key way to differentiate your business. RedRockResearch point out some excellent best practices for great IT/software support:
- Maintain Systems
- Manage Customer Expectations
- Become a Quick-Reaction Force
and as far as I can see, all three want excellent system monitoring and reporting.
Software Developers
Deploy your new / patched / modified application to a development server and run some load testing. How did it compare to last time? It might have been twice as fast delivering pages to your customers, but if it requires three times more RAM to achieve this, you probably want to know. But memory usage is fairly easy to predict.
What about how much time your cpu(s) spends in iowait when your app is running under load? Or do you get a lot more interrupts? What happens to disk IO? It is really easy to write a software patch that fixes a problem, but fail to notice that you’ve doubled the disk io requirements.
Some problems only appear when your traffic volumes get above some threshold, at which point you see “weirdness”.
Another trick is launching a new feature/service to live users (after suitable development testing) but limit the number that will see it. Aka, don’t put that new application on your homepage just yet. Invite a few users to try it, and while they do, keep a close eye on your server monitors and see if you cannot spot some things to improve.
Management
If your managers are doing their job, they will want to know how their systems are performing. Management dashboard anyone?
A good manager should soak up as much information as they can about what is going on with their team, company, environment. Your CFO might not care about the memory footprint of your webserver, but being able to justify to them why your twenty webservers need double the ram and faster disks is a lot easier when you can show them the graph about performance before/after you tried these changes on a single machine.
Tracking the number of support emails, orders, uptime etc. are quite likely to be things someone higher up the chain of command is sooner or later going to want to see.
Leave a comment