More system metrics

[jhofeditz]

We would like Telegraf to provide a few more system level metrics to fully replace our existing metrics collectors (Munin/Nagios/Collectd). It would be nice to see these added:

• total number of system processes/threads (by state - runnable,sleeping,zombie)
• interrupts and context switches
• netstat tcp metrics (established, reset, time wait, etc)
• NTP time offset
• more granular memory metrics (committed, active, cache/buffers, user/kernel)

[mabushey]

I would love:

• Top 10 (or however many configured) list of system processes
• Running kernel (uname -r)

[sparrc]

These all sound like reasonable requests to me. Currently I am focusing on statsd support but should have time to work on this later this month.

[yankcrime]

+1 on this. Would like to see more I/O-related stats using iostat perhaps, i.e queue depths.

[rudenkovk]

+1
We really need metrics: process count, threads count, threads per process

[rplessl]

+1, we will appreciate this too.

[titilambert]

+1 !
Which one should be the first ?
Processes state count ? like https://collectd.org/wiki/index.php/Plugin:Processes

[sparrc]

closing this because processes, kernel, and ntpq plugins are now in master, and will be available in 0.11.0, which covers all metrics in the original request.

@mabushey you could get one-off string metrics like that currently by making a simple script like this and putting it in the exec plugin:

echo kernel_version value="`uname -r`"

I would like to also have a simple "value" be a parsable input format in general, so that you could do this:

[[inputs.exec]]
  commands = ["uname -r"]
  data_format = "value"
  data_type = "string"

and parse a simple string, float, or int from any command.

[sparrc]

see #849

[whitelynx]

@mabushey wrote:

• Top 10 (or however many configured) list of system processes

@sparrc: Did support for this ever get implemented? I saw that #1932 was closed, but searching for top processes (or even just top) in the project hasn't shown me much other than that.

[douglasg14b]

How can we gather this information? I can't sort out if the top n feature was implemented somewhere?

[danielnelson]

@douglasg14b The top n feature was implemented as a processor, you can use it with the procstat input: https://github.com/influxdata/telegraf/tree/master/plugins/processors/topk