CHANGES.txt 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292
  1. 3.0a2
  2. - Fixed the README.txt example for defining the supervisor RPC
  3. interface in the configuration file. Thanks to Drew Perttula.
  4. - Fixed a bug where process communication events would not have the
  5. proper payload if the payload data was very short.
  6. - when supervisord attempted to kill a process with SIGKILL after
  7. the process was not killed within "stopwaitsecs" using a "normal"
  8. kill signal, supervisord would crash with an improper
  9. AssertionError. Thanks to Calvin Hendryx-Parker.
  10. - On Linux, Supervisor would consume too much CPU in an effective
  11. "busywait" between the time a subprocess exited and the time at
  12. which supervisor was notified of its exit status. Thanks to Drew
  13. Perttula.
  14. - RPC interface behavior change: if the RPC method
  15. "sendProcessStdin" is called against a process that has closed its
  16. stdin file descriptor (e.g. it has done the equivalent of
  17. "sys.stdin.close(); os.close(0)"), we return a NO_FILE fault
  18. instead of accepting the data.
  19. - Changed the semantics of the process configuration 'autorestart'
  20. parameter with respect to processes which move between the RUNNING
  21. and EXITED state. 'autorestart' was previously a boolean. Now
  22. it's a trinary, accepting one of 'false', 'unexpected', or 'true'.
  23. If it's 'false', a process will never be automatically restarted
  24. from the EXITED state. If it's 'unexpected', a process that
  25. enters the EXITED state will be automatically restarted if it
  26. exited with an exit code that was not named in the process
  27. config's 'exitcodes' list. If it's 'true', a process that enters
  28. the EXITED state will be automatically restarted unconditionally.
  29. The default is now 'unexpected' (it was previously 'true'). The
  30. readdition of this feature is a reversion of the behavior change
  31. note in the changelog notes for 3.0a1 that asserted we never cared
  32. about the process' exit status when determining whether to restart
  33. it or not.
  34. - setup.py develop (and presumably setup.py install) would fail
  35. under Python 2.3.3, because setuptools attempted to import
  36. 'splituser' from urllib2, and it didn't exist.
  37. - It's now possible to use 'setup.py install' and 'setup.py develop'
  38. on systems which do not have a C compiler if you set the environment
  39. variable "NO_MELD3_EXTENSION_MODULES=1" in the shell in which
  40. you invoke these commands (a new version of meld3, 0.6.1 was released
  41. which respects this envvar and does not try to compile its optional
  42. C extensions when it's set).
  43. - The test suite would fail on Python versions <= 2.3.3 because
  44. the "assertTrue" and "assertFalse" methods of unittest.TestCase
  45. didn't exist in those versions.
  46. - The 'supervisorctl' and 'supervisord' wrapper scripts were disused
  47. in favor of using setuptools' 'console_scripts' entry point settings.
  48. - Documentation files and the sample configuration file are put into
  49. the generated supervisor egg's 'doc' directory.
  50. 3.0a1
  51. - Default config file comment documented 10 secs as default for
  52. 'startsecs' value in process config, in reality it was 1 sec.
  53. Thanks to Christoph Zwerschke.
  54. - Make note of subprocess environment behavior in README.txt.
  55. Thanks to Christoph Zwerschke.
  56. - New "strip_ansi" config file option attempts to strip ANSI escape
  57. sequences from logs for smaller/more readable logs (submitted by
  58. Mike Naberezny).
  59. - The XML-RPC method supervisor.getVersion() has been renamed for
  60. clarity to supervisor.getAPIVersion(). The old name is aliased
  61. for compatibility but is deprecated and will be removed in a
  62. future version (Mike Naberezny).
  63. - Improved web interface styling (Mike Naberezny, Derek DeVries)
  64. - The XML-RPC method supervisor.startProcess() now checks that
  65. the file exists and is executable (Mike Naberezny).
  66. - Two environment variables, "SUPERVISOR_PROCESS_NAME" and
  67. "SUPERVISOR_PROCESS_GROUP" are set in the environment of child
  68. processes, representing the name of the process and group in
  69. supervisor's configuration.
  70. - Process state map change: a process may now move directly from the
  71. STARTING state to the STOPPING state (as a result of a stop
  72. request).
  73. - Behavior change: if 'autorestart' is true, even if a process exits
  74. with an "expected" exit code, it will still be restarted. In the
  75. immediately prior release of supervisor, this was true anyway, and
  76. no one complained, so we're going to consider that the "officially
  77. correct" behavior from now on.
  78. - Supervisor now logs subprocess stdout and stderr independently.
  79. The old program config keys "logfile", "logfile_backups" and
  80. "logfile_maxbytes" are superseded by "stdout_logfile",
  81. "stdout_logfile_backups", and "stdout_logfile_maxbytes". Added
  82. keys include "stderr_logfile", "stderr_logfile_backups", and
  83. "stderr_logfile_maxbytes". An additional "redirect_stderr" key is
  84. used to cause program stderr output to be sent to its stdin
  85. channel. The keys "log_stderr" and "log_stdout" have been
  86. removed.
  87. - '[program:x]' config file sections now represent "homgeneous
  88. process groups" instead of single processes. A "numprocs" key in
  89. the section represents the number of processes that are in the
  90. group. A "process_name" key in the section allows composition of
  91. the each process' name within the homogeneous group.
  92. - A new kind of config file section, '[group:x]' now exists,
  93. allowing users to group heterogeneous processes together into a
  94. process group that can be controlled as a unit from a client.
  95. - Supervisord now emits "events" at certain points in its normal
  96. operation. These events include supervisor state change events,
  97. process state change events, and "process communication events".
  98. - A new kind of config file section '[eventlistener:x]' now exists.
  99. Each section represents an "event listener pool", which is a
  100. special kind of homogeneous process group. Each process in the
  101. pool is meant to receive supervisor "events" via its stdin and
  102. perform some notification (e.g. send a mail, log, make an http
  103. request, etc.)
  104. - Supervisord can now capture data between special tokens in
  105. subprocess stdout/stderr output and emit a "process communications
  106. event" as a result.
  107. - Supervisor's XML-RPC interface may be extended arbitrarily by
  108. programmers. Additional top-level namespace XML-RPC interfaces
  109. can be added using the '[rpcinterface:foo]' declaration in the
  110. configuration file.
  111. - New 'supervisor'-namespace XML-RPC methods have been added:
  112. getAPIVersion (returns the XML-RPC API version, the older
  113. "getVersion" is now deprecated), "startProcessGroup" (starts all
  114. processes in a supervisor process group), "stopProcessGroup"
  115. (stops all processes in a supervisor process group), and
  116. "sendProcessStdin" (sends data to a process' stdin file
  117. descriptor).
  118. - 'supervisor'-namespace XML-RPC methods which previously accepted
  119. ony a process name as "name" (startProcess, stopProcess,
  120. getProcessInfo, readProcessLog, tailProcessLog, and
  121. clearProcessLog) now accept a "name" which may contain both the
  122. process name and the process group name in the form
  123. 'groupname:procname'. For backwards compatibility purposes,
  124. "simple" names will also be accepted but will be expanded
  125. internally (e.g. if "foo" is sent as a name, it will be expanded
  126. to "foo:foo", representing the foo process within the foo process
  127. group).
  128. - 2.X versions of supervisorctl will work against supervisor 3.0
  129. servers in a degraded fashion, but 3.X versions of supervisorctl
  130. will not work at all against supervisor 2.X servers.
  131. Known issues:
  132. - supervisorctl and the web interface do not yet allow you to stop
  133. / start / restart a process group as a unit.
  134. - supervisorctl and the web interface do not allow you to tail or
  135. otherwise examine stderr log files of processes.
  136. - buffered event notifications may be lost at supervisor shutdown
  137. or restart time.
  138. Acknowledgements:
  139. Maintainable Software (http://www.maintainable.com) contracted
  140. Agendless Consulting to add the event notification features and
  141. extensible XML-RPC namespaces feature to supervisor.
  142. 2.2b1
  143. - Individual program configuration sections can now specify an
  144. environment.
  145. - Added a 'version' command to supervisorctl. This returns the
  146. version of the supervisor2 package which the remote supervisord
  147. process is using.
  148. 2.1
  149. - When supervisord was invoked more than once, and its configuration
  150. was set up to use a UNIX domain socket as the HTTP server, the
  151. socket file would be erased in error. The symptom of this was
  152. that a subsequent invocation of supervisorctl could not find the
  153. socket file, so the process could not be controlled (it and all of
  154. its subprocesses would need to be killed by hand).
  155. - Close subprocess file descriptors properly when a subprocess exits
  156. or otherwise dies. This should result in fewer "too many open
  157. files to spawn foo" messages when supervisor is left up for long
  158. periods of time.
  159. - When a process was not killable with a "normal" signal at shutdown
  160. time, too many "INFO: waiting for x to die" messages would be sent
  161. to the log until we ended up killing the process with a SIGKILL.
  162. Now a maximum of one every three seconds is sent up until SIGKILL
  163. time. Thanks to Ian Bicking.
  164. - Add an assertion: we never want to try to marshal None to XML-RPC
  165. callers. Issue 223 in the collector from vgatto indicates that
  166. somehow a supervisor XML-RPC method is returning None (which
  167. should never happen), but I cannot identify how. Maybe the
  168. assertion will give us more clues if it happens again.
  169. - Supervisor would crash when run under Python 2.5 because the
  170. xmlrpclib.Transport class in Python 2.5 changed in a
  171. backward-incompatible way. Thanks to Eric Westra for the bug
  172. report and a fix.
  173. - Tests now pass under Python 2.5.
  174. - Better supervisorctl reporting on stop requests that have a FAILED
  175. status.
  176. - Removed duplicated code (readLog/readMainLog), thanks to Mike
  177. Naberezny.
  178. - Added tailProcessLog command to the XML-RPC API. It provides a
  179. more efficient way to tail logs than readProcessLog(). Use
  180. readProcessLog() to read chunks and tailProcessLog() to tail.
  181. (thanks to Mike Naberezny).
  182. 2.1b2
  183. - Added new tailProcessLog() command to the XML-RPC API that
  184. is more efficient for just tailing than the existing
  185. readProcessLog() command (Mike Naberezny).
  186. 2.1b1
  187. - "supervisord -h" and "supervisorctl -h" did not work (traceback
  188. instead of showing help view (thanks to Damjan from Macedonia for
  189. the bug report).
  190. - Processes which started successfully after failing to start
  191. initially are no longer reported in BACKOFF state once they are
  192. started successfully (thanks to Damjan from Macdonia for the bug
  193. report).
  194. - Add new 'maintail' command to supervisorctl shell, which allows
  195. you to tail the 'main' supervisor log. This uses a new
  196. readMainLog xmlrpc API.
  197. - Various process-state-transition related changes, all internal.
  198. README.txt updated with new state transition map.
  199. - startProcess and startAllProcesses xmlrpc APIs changed: instead of
  200. accepting a timeout integer, these accept a wait boolean (timeout
  201. is implied by process' "startsecs" configuration). If wait is
  202. False, do not wait for startsecs.
  203. Known issues:
  204. Code does not match state transition map. Processes which are
  205. configured as autorestarting which start "successfully" but
  206. subsequently die after 'startsecs' go through the transitions
  207. RUNNING -> BACKOFF -> STARTING instead of the correct transitions
  208. RUNNING -> EXITED -> STARTING. This has no real negative effect,
  209. but should be fixed for correctness.
  210. 2.0
  211. - pidfile written in daemon mode had incorrect pid.
  212. - supervisorctl: tail (non -f) did not pass through proper error
  213. messages when supplied by the server.
  214. - Log signal name used to kill processes at debug level.
  215. - supervisorctl "tail -f" didn't work with supervisorctl sections
  216. configured with an absolute unix:// URL
  217. - New "environment" config file option allows you to add environment
  218. variable values to supervisord environment from config file.
  219. 2.0b1
  220. - fundamental rewrite based on 1.0.6, use distutils (only) for
  221. installation, use ConfigParser rather than ZConfig, use HTTP for
  222. wire protocol, web interface, less lies in supervisorctl.