1

I just installed Munin for monitoring my web and database server. I lot of plugins were included out-of-the box, but I could use some advice on which one is actually good key indicators to monitor.

Disk IOs per device
Disk latency per device
Disk usage in percent
Inode usage in percent
Throughput per device
Utilization per device

Fork rate
Number of threads
Processes
Processes priority
VMstat

CPU usage
File table usage
Inode table usage
Load average
Memory usage
Swap in/out

So my question is; which of these parameters are key indicators and has any value to monitor?

Note: I also use the plugins for Apache and MySQL, but here I am interested in the disk/system/process plugins. I have asked advice for what to monitor on the database on Database Administrators: https://dba.stackexchange.com/questions/77162/what-to-monitor-key-indicators-for-mysql-server

Community
  • 1
Thomas Jensen
  • 133
  • 7

1 Answers1

1

Collect it all. You'll find a use for it later, or wish you had been collecting it from the beginning.

Of course, the things you aren't collecting here are the most important things. Is the database working? How well? Is the web server serving? Is the web site functioning? Can people still place orders?

Michael Hampton
  • 244,070
  • 43
  • 506
  • 972
  • I use Pingdom for monitoring the website availability and response time. And I use the Apache and MySql plugins for Munin, I have asked on Database Administrators what parameters I need to monitor for the database: http://dba.stackexchange.com/questions/77162/what-to-monitor-key-indicators-for-mysql-server – Thomas Jensen Sep 20 '14 at 20:11
  • Does Pingdom tell you, e.g., whether "My Cart" is giving certain users a 500 error? Munin? – Michael Hampton Sep 20 '14 at 20:12
  • Hm, no. It doesn't. I don't run a commercial website, but I do my best to give the users a good experience. How would I go about to test these things? – Thomas Jensen Sep 20 '14 at 20:17
  • For that you need something that will analyze your logs in (near) real time, such as logstash. – Michael Hampton Sep 20 '14 at 20:27