0

This picture shows the CPU Usage of a Debian 8 installation running on a Virtual Machine (VMWare ESXi). I'm wondering about two things here

  1. Why doesn't the usage sum up to 200% (the virtual machine has 2 Cores assigned)

  2. Why does the usage drops even more? Shouldn't be the missing percents be marked as idle?

The graph is from a munin installation.

CPU Load

HBruijn
  • 77,029
  • 24
  • 135
  • 201
0x4a42
  • 1

1 Answers1

3

Really you're asking "Why does Munin report things in a strange way?" There are a number of reasons for this to happen. RRDtool (the component that creates the graphs you can see) is not meant to be a one-to-one mapping of value at a given time, it is a graphical representation of a round-robin database. These databases average out data and they are a "lossy" way to store historical information. In other words, the same data looks different depending on the time range you specify.

The machine reporting the data might not be reporting data that sums up to 100% either.

Is this causing a problem? Are customers impacted? Otherwise I'd say there's nothing to worry about.

Joel E Salas
  • 5,572
  • 16
  • 25
  • Yes, I'm aware of the fact, that this graph is not very accurate. Still I'm wondering about it. And yes there is a problem for users because overall performance is dropping dramatically when uptime gets over about 14 days. The (at least for me) strange looking graph is a first hint to find the problem. – 0x4a42 Dec 18 '15 at 02:18
  • Does this graph capture when the problem starts? – Joel E Salas Dec 18 '15 at 19:19
  • Yes, it gets even worse then (CPU usage sums up to about 20% then). – 0x4a42 Dec 20 '15 at 14:36
  • @0x4a42 Honestly the high CPU might be stealing CPU time from the reporting component itself. You won't be able to use Munin alone to solve this problem. – Joel E Salas Dec 21 '15 at 04:30