README.rst 6.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219
  1. Welcome to the repository for Hue
  2. =================================
  3. .. note::
  4. This is the development-oriented readme. If you want to write notes for
  5. end users, please put them in ``dist/README``.
  6. Hue is both a web UI for Hadoop and a framework to create interactive web
  7. applications. It features a FileBrowser for accessing HDFS, JobSub and
  8. JobBrowser applications for submitting and viewing MapReduce jobs, a Beeswax
  9. application for interacting with Hive. On top of that, the web frontend
  10. is mostly built from declarative widgets that require no JavaScript and are
  11. easy to learn.
  12. File Layout
  13. ===========
  14. The "core" stuff is in ``desktop/core/``, whereas installable apps live in
  15. ``apps/``. Please place third-party dependencies in the app's ext-py/
  16. directory.
  17. The typical directory structure for inside an application includes:
  18. src/
  19. for Python code
  20. conf/
  21. for configuration (``.ini``) files to be installed
  22. static/
  23. for static HTML and js resources
  24. templates/
  25. for data to be put through a template engine
  26. docs/
  27. for helpful notes
  28. The python code is structured simply as
  29. ``module/package.py``,
  30. where module may be "filebrowser" or "jobsub". Because it is unlikely that
  31. there are going to be huge conflicts, we're going without a deep nested
  32. hierarchy.
  33. URL Layout
  34. ==========
  35. ``core/src/desktop/urls.py`` contains the current layout for top-level URLs.
  36. For the URLs within your application, you should make your own ``urls.py``
  37. which will be automatically rooted at ``/yourappname/`` in the global
  38. namespace. See ``apps/hello/src/hello/urls.py`` for an example.
  39. Development Prerequisites
  40. ===========================
  41. 1. On your host system, you need to have the python "virtualenv" package
  42. installed.
  43. 2. Also, you'll need these library development packages and tools installed on
  44. your system:
  45. Debian:
  46. * asciidoc
  47. * gcc
  48. * g++
  49. * libmysqlclient-dev
  50. * libsasl2-dev
  51. * libsqlite3-dev
  52. * libxml2-dev
  53. * libxslt-dev
  54. * maven2
  55. * python-dev
  56. * python-setuptools
  57. * python-simplejson
  58. CentOS:
  59. * asciidoc
  60. * cyrus-sasl-devel
  61. * gcc
  62. * libxml2-devel
  63. * libxslt-devel
  64. * maven2
  65. * mysql
  66. * mysql-devel
  67. * python-devel
  68. * python-setuptools
  69. * python-simplejson (for the crepo tool)
  70. * sqlite-devel
  71. MacOS (mac port):
  72. * liblxml
  73. * libxml2
  74. * libxslt
  75. * mysql5-devel
  76. * simplejson (easy_install)
  77. * sqlite3
  78. 3. You need to have crepo installed, and preferably on your path. If it is not
  79. on your path, set the environment variable ``CREPO`` to point to ``crepo.py``
  80. from that distribution. You can clone crepo from
  81. http://github.com/cloudera/crepo.git somewhere else on your system.
  82. Getting Started
  83. ===============
  84. To build and get the core server running::
  85. $ export HADOOP_HOME=<path-to-hadoop-home>
  86. $ git clone http://github.com/cloudera/hue.git
  87. $ cd hue
  88. $ make apps
  89. $ build/env/bin/hue runserver_plus
  90. To start the helper daemons::
  91. $ build/env/bin/hue beeswax_server
  92. $ build/env/bin/hue jobsubd
  93. Now Hue should be running on http://localhost:8000.
  94. Setting up Hadoop
  95. =================
  96. In order to start up a pseudo-distributed cluster with the plugins enabled,
  97. run::
  98. $ ./tools/scripts/configure-hadoop.sh all
  99. After doing so, running ``jps`` should show all the daemons running (NN, JT,
  100. TT, DN) and you should be able to see the web UI on http://localhost:50030/ and
  101. http://localhost:50070/.
  102. FAQ
  103. ===
  104. 1: What does "Exception: no app!" mean?
  105. Your template has an error in it. Check for messages from the server that
  106. look like::
  107. INFO:root:Processing exception: Unclosed tag 'if'. Looking for one of: else, endif
  108. 2: What do I do if I get "There was an error launching ..."?
  109. Turn on debugging by issuing ``dbug.cookie()`` in a Firebug console.
  110. Django Conventions
  111. ==================
  112. If you need to name your urls
  113. (http://docs.djangoproject.com/en/dev/topics/http/urls/#naming-url-patterns)
  114. because there's ambiguity in the view, be sure to prefix the name
  115. with the application name. The url name namespace is global. So
  116. ``jobsub.list`` is fine, but ``list`` is not.
  117. Hue is using Django 1.1, which supports the notion of URL namespaces:
  118. http://docs.djangoproject.com/en/dev/topics/http/urls/#url-namespaces.
  119. We have yet to move over our URLs to this construct. Brownie points for the
  120. developer who takes this on.
  121. Using and Installing Thrift
  122. ===========================
  123. Right now, we check in the generated thrift code.
  124. To generate the code, you'll need the thrift binary.
  125. Compile it like so::
  126. $ git clone http://github.com/dreiss/thrift.git
  127. $ cd thrift
  128. $ ./bootstrap.sh
  129. $ ./configure --with-py=no --with-java=no --with-perl=no --prefix=$HOME/pub
  130. We exclude python, java, and perl because they don't like
  131. to install in prefix. If you look around at configure's --help,
  132. there are environment variables that determine where those
  133. runtime bindings are installed.
  134. ::
  135. $ make && make install
  136. When preparing ``.thrift`` files, you can use she-bangs to generate
  137. the python bindings like so::
  138. #!/usr/bin/env thrift -r --gen py:new_style -o ../../../
  139. .. note::
  140. This file is in reStructuredText. You may run
  141. ``rst2html README.rst > README.html`` to produce a HTML.
  142. Profiling Hue Apps
  143. ==================
  144. Hue has a profiling system built in, which can be used to analyze server-side
  145. performance of applications. To enable profiling::
  146. $ build/env/bin/hue runprofileserver
  147. Then, access the page that you want to profile. This will create files like
  148. /tmp/useradmin.users.000072ms.2011-02-21T13:03:39.745851.prof. The format for
  149. the file names is /tmp/<app_module>.<page_url>.<time_taken>.<timestamp>.prof.
  150. Hue uses the hotshot profiling library for instrumentation. The documentation
  151. for this library is located at: http://docs.python.org/library/hotshot.html.
  152. To make use of the profiling data quickly, you can create a script that does
  153. the following::
  154. #!/usr/bin/python
  155. import hotshot.stats
  156. import sys
  157. stats = hotshot.stats.load(sys.argv[1])
  158. stats.sort_stats('cumulative', 'calls')
  159. stats.print_stats(100)
  160. This script takes in a .prof file, and orders function calls by the cumulative
  161. time spent in that function, followed by the number of times the function was
  162. called, and then prints out the top 100 time-wasters. For information on the
  163. other stats available, take a look at this website:
  164. http://docs.python.org/library/profile.html#pstats.Stats