NAME

vserver_cpu_ - Wildcard plugin to graph CPU usage types per vserver

CONFIGURATION

 [vserver_cpu_*]
   env.vservers - List of vservers to include in the graph, or "all"

DEFAULT CONFIGURATION

 [vserver_cpu_*]
   env.vservers all

USAGE

Configuration variables vservers - specify the vservers to include in the graph (default: all)

NOTE: If no configuration variable is set, the default will be used

see vserver_resources for example uses of configuration files

or links to define what to monitor: vserver_cpu_ -> monitor cpu usage of all vservers on all cpus vserver_hold_ -> monitor hold on all vservers on all cpus vserver_hold_0 -> monitor hold on all vservers on cpu0 vserver_hold_1 -> monitor hold on all vservers on cpu1 vserver_hold_foo -> monitor hold on all cpus on vserver named foo vserver_sys_foo -> monitor cpu usage on all cpus on vserver named foo

AUTHOR

Copyright (C) 2006-2008 Holger Levsen and Micah Anderson

LICENSE

GNU GPLv2

TODO

Less shell subprocesses

The plugin uses lots and lots of shell subprocesses. Maybe factor common vserver plugin code into Munin::Plugin::Linux::Vserver?

Comments

Comment the code or go mad

Jiffies per second

Add info how many jiffies per second are available on a machine

Split CPU time

User and system cpu are always added to each other, make it optional to split them?

Less use of /proc

use /proc less often (100 times more overhead than talking to the kernel directly) i.e. use something like pagesize=\`perl -MPOSIX -e 'print POSIX::sysconf(_SC_PAGESIZE), "\n";'\`

(Accoding to many using /proc is a feature not a bug - Nicoali Langfeldt 2009-12-23)

MAGIC MARKERS

  #%# family=auto
  #%# capabilities=autoconf suggest