README.rst 6.4 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217
  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. Also, you'll need these library development packages and tools installed on
  42. your system:
  43. Debian:
  44. * ant
  45. * asciidoc
  46. * gcc
  47. * g++
  48. * libmysqlclient-dev
  49. * libsasl2-dev
  50. * libsqlite3-dev
  51. * libxml2-dev
  52. * libxslt-dev
  53. * maven2
  54. * python-dev
  55. * python-simplejson
  56. CentOS:
  57. * ant
  58. * asciidoc
  59. * cyrus-sasl-devel
  60. * gcc
  61. * gcc-c++
  62. * libxml2-devel
  63. * libxslt-devel
  64. * maven2
  65. * mysql
  66. * mysql-devel
  67. * python-devel
  68. * python-simplejson (for the crepo tool)
  69. * sqlite-devel
  70. MacOS (mac port):
  71. * liblxml
  72. * libxml2
  73. * libxslt
  74. * mysql5-devel
  75. * simplejson (easy_install)
  76. * sqlite3
  77. 2. You need to have crepo installed, and preferably on your path. If it is not
  78. on your path, set the environment variable ``CREPO`` to point to ``crepo.py``
  79. from that distribution. You can clone crepo from
  80. http://github.com/cloudera/crepo.git somewhere else on your system.
  81. Getting Started
  82. ===============
  83. To build and get the core server running::
  84. $ export HADOOP_HOME=<path-to-hadoop-home>
  85. $ git clone http://github.com/cloudera/hue.git
  86. $ cd hue
  87. $ make apps
  88. $ build/env/bin/hue runserver_plus
  89. To start the helper daemons::
  90. $ build/env/bin/hue beeswax_server
  91. $ build/env/bin/hue jobsubd
  92. Now Hue should be running on http://localhost:8000.
  93. Setting up Hadoop
  94. =================
  95. In order to start up a pseudo-distributed cluster with the plugins enabled,
  96. run::
  97. $ ./tools/scripts/configure-hadoop.sh all
  98. After doing so, running ``jps`` should show all the daemons running (NN, JT,
  99. TT, DN) and you should be able to see the web UI on http://localhost:50030/ and
  100. http://localhost:50070/.
  101. FAQ
  102. ===
  103. 1: What does "Exception: no app!" mean?
  104. Your template has an error in it. Check for messages from the server that
  105. look like::
  106. INFO:root:Processing exception: Unclosed tag 'if'. Looking for one of: else, endif
  107. 2: What do I do if I get "There was an error launching ..."?
  108. Turn on debugging by issuing ``dbug.cookie()`` in a Firebug console.
  109. Django Conventions
  110. ==================
  111. If you need to name your urls
  112. (http://docs.djangoproject.com/en/dev/topics/http/urls/#naming-url-patterns)
  113. because there's ambiguity in the view, be sure to prefix the name
  114. with the application name. The url name namespace is global. So
  115. ``jobsub.list`` is fine, but ``list`` is not.
  116. Hue is using Django 1.1, which supports the notion of URL namespaces:
  117. http://docs.djangoproject.com/en/dev/topics/http/urls/#url-namespaces.
  118. We have yet to move over our URLs to this construct. Brownie points for the
  119. developer who takes this on.
  120. Using and Installing Thrift
  121. ===========================
  122. Right now, we check in the generated thrift code.
  123. To generate the code, you'll need the thrift binary.
  124. Compile it like so::
  125. $ git clone http://github.com/dreiss/thrift.git
  126. $ cd thrift
  127. $ ./bootstrap.sh
  128. $ ./configure --with-py=no --with-java=no --with-perl=no --prefix=$HOME/pub
  129. We exclude python, java, and perl because they don't like
  130. to install in prefix. If you look around at configure's --help,
  131. there are environment variables that determine where those
  132. runtime bindings are installed.
  133. ::
  134. $ make && make install
  135. When preparing ``.thrift`` files, you can use she-bangs to generate
  136. the python bindings like so::
  137. #!/usr/bin/env thrift -r --gen py:new_style -o ../../../
  138. .. note::
  139. This file is in reStructuredText. You may run
  140. ``rst2html README.rst > README.html`` to produce a HTML.
  141. Profiling Hue Apps
  142. ==================
  143. Hue has a profiling system built in, which can be used to analyze server-side
  144. performance of applications. To enable profiling::
  145. $ build/env/bin/hue runprofileserver
  146. Then, access the page that you want to profile. This will create files like
  147. /tmp/useradmin.users.000072ms.2011-02-21T13:03:39.745851.prof. The format for
  148. the file names is /tmp/<app_module>.<page_url>.<time_taken>.<timestamp>.prof.
  149. Hue uses the hotshot profiling library for instrumentation. The documentation
  150. for this library is located at: http://docs.python.org/library/hotshot.html.
  151. To make use of the profiling data quickly, you can create a script that does
  152. the following::
  153. #!/usr/bin/python
  154. import hotshot.stats
  155. import sys
  156. stats = hotshot.stats.load(sys.argv[1])
  157. stats.sort_stats('cumulative', 'calls')
  158. stats.print_stats(100)
  159. This script takes in a .prof file, and orders function calls by the cumulative
  160. time spent in that function, followed by the number of times the function was
  161. called, and then prints out the top 100 time-wasters. For information on the
  162. other stats available, take a look at this website:
  163. http://docs.python.org/library/profile.html#pstats.Stats