Difference between revisions of "Cgroups"

From ArchWiki
Jump to: navigation, search
(Systemd service file instead of old init script.)
(Ad-hoc groups: cgroup renamed from "me" to "groupname".)
Line 17: Line 17:
 
== Simple usage ==
 
== Simple usage ==
 
=== Ad-hoc groups ===
 
=== Ad-hoc groups ===
One of the powers of cgroups is that you can create "ad-hoc" groups on the fly. In fact, you can even grant the privileges to create custom groups to regular users. Run this as root (replace $USER with your user name):
+
One of the powers of cgroups is that you can create "ad-hoc" groups on the fly. In fact, you can even grant the privileges to create custom groups to regular users. Run this as root (replace $USER with your user name and '''groupname''' with the name you want to give to the cgroup):
  
  sudo '''cgcreate''' -a '''$USER''' -g memory,cpu:'''me'''
+
  sudo '''cgcreate''' -a '''$USER''' -g memory,cpu:'''groupname'''
  
That's it! Now all the tunables in the group "me" are writable by your user:
+
That's it! Now all the tunables in the group '''groupname''' are writable by your user:
  
  $ ls -l /mnt/cgroups/memory/<b>me</b>
+
  $ ls -l /mnt/cgroups/memory/<b>groupname</b>
 
  total 0
 
  total 0
 
  -rwxrwxr-x 1 user root 0 Sep 25 00:39 cgroup.event_control
 
  -rwxrwxr-x 1 user root 0 Sep 25 00:39 cgroup.event_control
Line 35: Line 35:
 
Cgroups are hierarchical, so you can create as many subgroups as you like. Let's say that, as a normal user, you want to run a '''bash''' shell under a new subgroup called 'foo':
 
Cgroups are hierarchical, so you can create as many subgroups as you like. Let's say that, as a normal user, you want to run a '''bash''' shell under a new subgroup called 'foo':
  
  cgcreate -g memory,cpu:'''me/foo'''
+
  cgcreate -g memory,cpu:'''groupname/foo'''
  '''cgexec'''  -g memory,cpu:me/foo '''bash'''
+
  '''cgexec'''  -g memory,cpu:groupname/foo '''bash'''
  
 
There we go! Just to make sure:
 
There we go! Just to make sure:
  
 
  $ cat /proc/self/cgroup
 
  $ cat /proc/self/cgroup
  11:memory:/me/foo
+
  11:memory:/groupname/foo
  6:cpu:/me/foo
+
  6:cpu:/groupname/foo
  
 
A new subdirectory was created for this group. To limit the memory usage of all processes in this group to '''10 MB''', run the following:
 
A new subdirectory was created for this group. To limit the memory usage of all processes in this group to '''10 MB''', run the following:
  
  $ echo <b>10000000</b> > /mnt/cgroups/memory/me/foo/<b>memory.limit_in_bytes</b>
+
  $ echo <b>10000000</b> > /mnt/cgroups/memory/groupname/foo/<b>memory.limit_in_bytes</b>
  
 
Note that the memory limit applies to RAM use only -- once tasks hit this limit, they will begin to swap. But it won't affect the performance of other processes significantly.
 
Note that the memory limit applies to RAM use only -- once tasks hit this limit, they will begin to swap. But it won't affect the performance of other processes significantly.
Line 52: Line 52:
 
Similarly you can change the CPU priority ("shares") of this group. By default all groups have '''1024''' shares. A group with '''100''' shares will get a ~10% portion of the CPU time:
 
Similarly you can change the CPU priority ("shares") of this group. By default all groups have '''1024''' shares. A group with '''100''' shares will get a ~10% portion of the CPU time:
  
  $ echo <b>100</b> > /mnt/cgroups/cpu/me/foo/<b>cpu.shares</b>
+
  $ echo <b>100</b> > /mnt/cgroups/cpu/groupname/foo/<b>cpu.shares</b>
  
 
You can find more tunables or statistics by listing the cgroup directory.
 
You can find more tunables or statistics by listing the cgroup directory.
Line 60: Line 60:
 
  $ pidof bash
 
  $ pidof bash
 
  13244 13266
 
  13244 13266
  $ '''cgclassify''' -g memory,cpu:me/foo `pidof bash`
+
  $ '''cgclassify''' -g memory,cpu:groupname/foo `pidof bash`
 
  $ cat /proc/13244/cgroup
 
  $ cat /proc/13244/cgroup
  11:memory:/me/foo
+
  11:memory:/groupname/foo
  6:cpu:/me/foo
+
  6:cpu:/groupname/foo
  
 
=== Persistent group configuration ===
 
=== Persistent group configuration ===

Revision as of 19:25, 15 January 2013

cgroups (aka control groups) is a Linux kernel feature to limit, police and account the resource usage of certain processes (actually process groups). Compared to other approaches like the 'nice' command or /etc/security/limits.conf, cgroups are infinitely more flexible.

Control groups can be used in multiple ways:

  • create and manage them on the fly using tools like cgcreate, cgexec, cgclassify etc
  • the "rules engine daemon", to automatically move certain users/groups/commands to groups (/etc/cgrules.conf and /usr/lib/systemd/system/cgconfig.service)
  • through other software such as Linux Containers (LXC) virtualization

Unfortunately this feature is often underappreciated due to lack of easy "how-to" style documentation. This is an attempt of fixing the problem. :)

Note: I'm only learning to use cgroups as I type this, so do not take everything I say here as pure gold.

Installing

First, install the utilities for managing cgroups; you need to install the libcgroupAUR package from the AUR.

Simple usage

Ad-hoc groups

One of the powers of cgroups is that you can create "ad-hoc" groups on the fly. In fact, you can even grant the privileges to create custom groups to regular users. Run this as root (replace $USER with your user name and groupname with the name you want to give to the cgroup):

sudo cgcreate -a $USER -g memory,cpu:groupname

That's it! Now all the tunables in the group groupname are writable by your user:

$ ls -l /mnt/cgroups/memory/groupname
total 0
-rwxrwxr-x 1 user root 0 Sep 25 00:39 cgroup.event_control
-rwxrwxr-x 1 user root 0 Sep 25 00:39 cgroup.procs
-rwxrwxr-x 1 user root 0 Sep 25 00:39 cpu.rt_period_us
-rwxrwxr-x 1 user root 0 Sep 25 00:39 cpu.rt_runtime_us
-rwxrwxr-x 1 user root 0 Sep 25 00:39 cpu.shares
-rwxrwxr-x 1 user root 0 Sep 25 00:39 notify_on_release
-rwxrwxr-x 1 user root 0 Sep 25 00:39 tasks

Cgroups are hierarchical, so you can create as many subgroups as you like. Let's say that, as a normal user, you want to run a bash shell under a new subgroup called 'foo':

cgcreate -g memory,cpu:groupname/foo
cgexec   -g memory,cpu:groupname/foo bash

There we go! Just to make sure:

$ cat /proc/self/cgroup
11:memory:/groupname/foo
6:cpu:/groupname/foo

A new subdirectory was created for this group. To limit the memory usage of all processes in this group to 10 MB, run the following:

$ echo 10000000 > /mnt/cgroups/memory/groupname/foo/memory.limit_in_bytes

Note that the memory limit applies to RAM use only -- once tasks hit this limit, they will begin to swap. But it won't affect the performance of other processes significantly.

Similarly you can change the CPU priority ("shares") of this group. By default all groups have 1024 shares. A group with 100 shares will get a ~10% portion of the CPU time:

$ echo 100 > /mnt/cgroups/cpu/groupname/foo/cpu.shares

You can find more tunables or statistics by listing the cgroup directory.

You can also change the cgroup of already running processes. To move all 'bash' commands to this group:

$ pidof bash
13244 13266
$ cgclassify -g memory,cpu:groupname/foo `pidof bash`
$ cat /proc/13244/cgroup
11:memory:/groupname/foo
6:cpu:/groupname/foo

Persistent group configuration

If you want your cgroups to be created at boot, you can define them in /etc/cgconfig.conf instead. For example, the "me" and "me/foo" group definitions would look like this:

/etc/cgconfig.conf 
group me {
  perm {
# who can manage limits
    admin {
      uid = $USER;
    }
# who can add tasks to this group
    task {
      uid = $USER;
    }
  }
# create this group in cpu and memory controllers
  cpu { }
  memory { }
}

group me/foo {
  cpu {
    cpu.shares = 100;
  }
  memory {
    memory.limit_in_bytes = 10000000;
  }
}

Documentation

For information on controllers and what certain switches and tunables mean, refer to kernel's Documentation/cgroup (or install linux-docs and see /usr/src/linux/Documentation/cgroup

For commands and configuration files, see relevant man pages, e.g. man cgcreate or man cgrules.conf