Skip to content

Latest commit

 

History

History
209 lines (164 loc) · 3.73 KB

README.md

File metadata and controls

209 lines (164 loc) · 3.73 KB

Usage

pip install flask tornado simplejson
python server/main.py

Core / Engine

We find that in the real world every data-source can be represented as a basic structure, a TREE of all types of data, and the tree is efficiency for communication between sub-systems. Based on this idea, we start to design a new realtime monitor system, which contains a HUGE tree in memroy, for realtime usage. History trend data is store in databases.

Here is an example of such structure:

[
{'cpu': {'10.3.1.12': {'idle': 32,
                       'sytem': 29,
                       'user': 900,
                       'marker': 'sth. wrong happened'}},
  'defaultaction': 'average'},
{'br':  {'ie':  {'6': 3},
                {'7': 8, {'venders':{'sogou':2}}},
                {'7' : {'venders':{'sogou':1}}}},
      {'webkit': {'14.0': 40},
                {'13.0': 20}},
  'defaultaction': 'add'}
]

Query by key name as form like:

  • br,webkit

  • br,ie,,venders,sogou

More detail in server/tree.py

server

directory layout

server/
  main.py
  tree.py -> async store in trend db
  treedb.py
  pushviahttp.py
  pushviasocket.py

server/pipes/*.py
server/web
  web/pages.py
  web/trend.py
server/web/static/*.js *.css *.html
server/web/templates/*.html
server/graph
server/graph/styles.py
  cpu|net|url|default
  input: tree
    bar|line|x-axis|marker range,picture|theme...|format
  output: png|pdf


# relay server
--relay
? agent/conf/relay.conf

url map

request argument name style like: e or a,b or a,,c

/api
  /api/q/a,b.json current tree
  

/push
  form field support json/accesslog
  json -> merge into core tree
  accesslog -> pipes -> tree -> merge into core tree

/pull/?key-br
/pull/?key-br,ie
/pull/?key-br,ie,7
  comet, repsonse in json list
    [{br,ie,7 : 2328},
    {br,ie,sogou,3 : 323}]
/get/
  response immediatly in json list
/realtime/
  page, init with /get/
  then comet by /pull/
/realtime/view/
  view by name, per user's specify

/trend/view/
  [default in one year]
  left tree | right (multi) js graph (zoomable highcharts/examples/dynamic-master-detail.htm)
/trend/view/?
  one js graph
  add compare | zoomable | setting style ...
/trend/?time
/setting/view
  add | remove view


/diff/keya/keyb
/graph/[.jpg|.png|.svg]

pipes

web access log
  browser type  -> tree
  os type       -> tree
  upstream time  -> tree
  request time
  response bytes
  url count
  error count
  ? geo

write your own pipe.py
  def init()
  def process(arr)
  def cleanup()

running mode:
init:
  for m in pipes/*.py:
    pipes.push(m)
    m.init()

run:
  for line in log.read():
    arr - weblog.parse(line)
    for m in pipes:
      m.process(arr)	

server push

  • udp

  • tcp

  • http post in form

Agent

directory

agent/
agent/modules/*.py *.so  
             gmod[32|64]/support ganglia module
agent/conf/*.conf *.on *.off
agent/agent.conf

agent/setup.py

modules

web access log(raw)
url(time to first byte, status code, bytes, ...)
  nginx stats

write your own module(same as [ganglia python module](http://sourceforge.net/apps/trac/ganglia/wiki/ganglia_gmond_python_modules#WritingcustomPythonmodules)):
  def metric_init({})
  def metric_cleanup()

reminder

A: use ganglia-plugin. build a package named 'ganglia-plugin' ? use ganglia-plugin only

  1. LD_PRELOAD=libganglia.so python modload.so modcpu.so

  2. rebuild modxxx.so

  3. ganglia install path/*.so

A: PUT/POST, socket

x post file in form ? HTTP POST raw body ? HTTP PUT socket tornado socket libevent socket + flask x web socket

A: 1 web log => graph