cgrouparch

Table of Content

• Name
• Status
• Description
• Synopsis
• Methods
  • manager.run
  • get_cgexec_arg
• Read cgroup usage info data
• Author
• Copyright and License

Name

cgrouparch

A python lib used to build cgroup directory tree, add set cgroup pid.

Status

This library is considered production ready.

Description

This lib is used to set up cgroup directory tree according to configuration saved in zookeeper, and add pid to cgroup accordingly.

Synopsis

Start the manager, the manager will watch on conf in zookeeper, rebuild cgroup directory tree if conf changed, it also update cgroup tasks file periodically according to conf.

# configuration in zookeeper:
{
    'cpu': {
        'sub_cgroup': {
            'test_cgroup_a': {
                'conf': {
                    'share': 1024,
                },
            },
            'test_cgroup_b': {
                'conf': {
                    'share': 100,
                },
                'sub_cgroup': {
                    'test_cgroup_b_sub1': {
                        'conf': {
                            'share': 200,
                        },
                    },
                },
            },
        },
    },
}

from pykit.cgrouparch import manager

def get_cgroup_pid_file(cgroup_name):
    if cgroup_name == 'test_cgroup_a':
        return ['/tmp/test.pid']
    # ...

def get_zk_host():
    return '127.0.0.1:2181,1.2.3.4:2181'

argkv = {
    'cgroup_dir': '/sys/fs/cgroup',
    'get_cgroup_pid_file': get_cgroup_pid_file,
    'get_zk_host': get_zk_host,
    'zk_prefix': '/cluser_a/service_rank',
    'zk_auth_data': [('digest', 'super:123456')],
    'communicate_ip': '127.0.0.1',
    'communicate_port': '3344',
}

manager.run(**argkv)

Get cgexec argument, to get the argument(for example: '-g cpu:test_cgroup_b/test_cgroup_b_sub1') used in cgexec command.

This is usefull if you want to use cgexec command to run you process in a cgroup, so the pid of process is added to the cgroup immediately.

You need to provide info about zookeeper, because we need to read the conf in zookeeper.

from pykit.cgrouparch import manager

argkv = {
    'cgroup_dir': '/sys/fs/cgroup',
    'get_zk_host': get_zk_host,
    'zk_prefix': '/cluser_a/service_rank',
    'zk_auth_data': [('digest', 'super:123456')],
}
cgexec_arg = manager.get_cgexec_arg(['test_cgroup_a'], **argkv)

# return like:
# {
#     'test_cgroup_a': '-g cpu:test_cgroup_a',
# }

Methods

manager.run

syntax: manager.run(**argkv)

This function read configuration from zookeeper and build the cgroup directory tree accordingly, it also update cgroup pid periodically. Every second it save usage info of each cgroup in redis, you can read usage info through websocket protocol.

arguments:

• get_cgroup_pid_file: a callback function, the argument is the cgroup name, and it should return a list of pid files. Required.

• get_zk_host: a callback function, should return zookeeper hosts, for example: '127.0.0.1:2181,1.2.3.4:2181'. Required.

• zk_prefix: specify the zookeeper key prefix'. Required.

• zk_auth_data: specify zookeeper auth data, for example: [('digest', 'super:123456')]. Required.

• communicate_ip: specify ip address the websocket server will bind. Default to '0.0.0.0'.

• communicate_port: specify port the websocket server will bind. Default to 43409.

• tasks_update_interval: set cgroup tasks file update interval in seconds. Default to 30.

• cgroup_dir: the mount point of the cgroup filesystem, default to '/sys/fs/cgroup'.

• redis_ip: ip of the redis server. Required.

• redis_port: port of the redis server. Required.

• redis_prefix: the key prefix to use when saving usage info into redis.

• redis_expire_time: we only keep recent usage info in redis, this specify the expire time in seconds of usage info data in redis. Default to 300.

• protected_cgroup: specify a list of cgroup names that you do not allow the manager to touch, or the manager will remove all cgroups that are not in the conf. Optional.

return Not return

get_cgexec_arg

syntax: manager.get_cgexec_arg(cgroup_names, **argkv)

This method used to get the argument used in cgexec command.

arguments:

• cgroup_names: a list of cgroup names.

• argkv: the same as that in manager.run, but only cgroup_dir, get_zk_host, zk_prefix, zk_auth_data are needed.

return

A dict contains argument for cgexec of each input cgroup.

Read cgroup usage info data

You can read usage data of each cgroup of each second through websocket protocol.

import json
import websocket

body = {
    'cmd': 'show_account',
    'args': {
        'nr_slot': 2,
    },
}

ws = websocket.WebSocket()
ws.connect('ws://127.0.0.1:43409')
ws.send(json.dumps(body))
print ws.recv()

# output example:
{
    "1525258774": {
        "cpu": {
            "value": {
                "usage": 1446489231295
            },
            "sub_cgroup": {
                "test_cgroup_a": {
                    "value": {
                        "usage": 192790540163
                    },
                },
                "test_cgroup_b": {
                    "value": {
                        "usage": 192790540163
                    },
                    "sub_cgroup": {
                        "test_cgroup_b_sub1": {
                            "value": {
                                "usage": 3928774282
                            },
                        },
                    },
                },
            },
        },
        "blkio": {
            ...
        },
    },
    "1525258775": {
        ...
    }
}

Author

Renzhi (任稚) [email protected]

Copyright and License

The MIT License (MIT)

Copyright (c) 2016 Renzhi (任稚) [email protected]