Bandwidth Graphs, CPU Percents?

[Fred Tyler]

I feel like I'm missing something really obvious, but I can't figure out how to get bandwidth graphs similar to those generated by MRTG, nor do I see stats on CPU utilization.

Everything else seems fine -- the Zabbix server is running, data is being collected from the machines running the Zabbix client, I get data on free disk space, server load, etc, but I don't see anything about bandwidth or CPU utilization.

I see screenshots of these things in the Zabbix documentation, so I guess other people are seeing them. Any suggestions?

(For the record, this is using Zabbix 1.1.7 and the Unix_t templates. The machines running Zabbix client are all Linux boxes.)

[paultorben]

I'm using Zabbix 1.1.6 -

For the Graphs - two suggestions:
- click on the "monitoring" tab, then click "latest data" - there's a 'graph' button there on the right in the "history" column.
- under "configuration" there's a 'graph' button - click that, then click on the "create Graph" button - lots of ways to make graphs here


For the CPU data - it is called "Load" - do you have data for that value coming in?

[Fred Tyler]

I'm familiar with the graphs; the problem is that none of the graphs I see correspond to network traffic. The closest thing I see is something called "Input statistic on interface eth0", but that appears to be some kind of cumulative statistic, as the current value for it is something like 37,000,000, and it just keeps going up.

And I do have a report for "load", but that's not quite the same thing as CPU utilization. For example, when there's a lot of disk activity on slower hard drives, you can see a really high load (10+) but be see very small CPU utilization (<10%). Or conversely, a single-threaded app can easily eat up 100% of the CPU for an hour and you might never see your load go much above 1. So I would like to know how I can get a reading of both.

I'm trying to migrate away from Cacti at the moment, and both of these statistics were enabled by default on Cacti. I suppose I could live without the CPU usage, but the bandwidth graphs are pretty much critical.

[Fred Tyler]

Ah, ok, by defining a new graph based on the average of "Input statistic on eth0" I was able to re-create what I believe is a bandwidth chart. So I think that problem is solved (thanks Paul -- I hadn't tried creating my own graphs until I read your comment).

That only leaves CPU utilization. You can see the CPU utilization when you run 'top' in Unix, or look at the task manager in Windows. Each process has a number listed in the "%CPU" column, and (on Linux, at least) at the top of the screen you see the total CPU usage for the system. Maybe Zabbix doesn't support this?

[rajan.s.ravi]

Hi,

You can definitely query %CPU. In Linux you can use system.cpu.util[,,avg1] and windows you can get it by defining perfmon parameters in the config file.

Ravi

[bbrendon]

Hi,

You can definitely query %CPU. In Linux you can use system.cpu.util[,,avg1] and windows you can get it by defining perfmon parameters in the config file.

That gets you user CPU usage. AFAIK there is no way in linux to get total CPU usage with the zabbix agent builtins.