9 from htsworkflow.util import mount
13 from pyinotify import EventsCodes
15 from benderjab import rpc
17 def is_runfolder(name):
21 >>> print is_runfolder('090630_HWUSI-EAS999_0006_30LNFAAXX')
23 >>> print is_runfolder('hello')
26 if re.match("[0-9]{6}_.*", name):
31 def get_top_dir(root, path):
33 Return the directory in path that is a subdirectory of root.
36 >>> print get_top_dir('/a/b/c', '/a/b/c/d/e/f')
38 >>> print get_top_dir('/a/b/c/', '/a/b/c/d/e/f')
40 >>> print get_top_dir('/a/b/c', '/g/e/f')
42 >>> print get_top_dir('/a/b/c', '/a/b/c')
45 if path.startswith(root):
46 subpath = path[len(root):]
47 if subpath.startswith('/'):
49 return subpath.split(os.path.sep)[0]
53 class WatcherEvent(object):
55 Track information about a file event
57 Currently its time, and if it was an indication we've completed the run.
59 def __init__(self, event_root=None):
60 self.time = time.time()
61 self.event_root = event_root
64 def __unicode__(self):
66 complete = "(completed)"
69 return u"<WatchEvent: %s %s %s>" % (time.ctime(self.time), self.event_root, complete)
71 class Handler(pyinotify.ProcessEvent):
72 def __init__(self, watchmanager, bot, completion_files=None):
74 Completion file contains current "completion" filename
77 self.watchmanager = watchmanager
79 if completion_files is not None:
80 completion_files = [ x.lower() for x in completion_files ]
81 self.completion_files = completion_files
83 def process_IN_CREATE(self, event):
84 for wdd in self.bot.wdds:
85 for watch_path in self.bot.watchdirs:
86 run_already_complete = False
87 # I only care about things created inside the watch directory, so
88 # the event path needs to be longer than the watch path in addition to
89 # starting with the watch_path
90 if len(event.path) > len(watch_path) and event.path.startswith(watch_path):
91 # compute name of the top level directory that had an event
92 # in the current watch path
93 target = get_top_dir(watch_path, event.path)
94 runfolder = os.path.join(watch_path, target)
96 if not is_runfolder(target):
97 logging.debug("Skipping %s, not a runfolder" % (target,))
100 # grab the previous events for this watch path
101 watch_path_events = self.last_event.setdefault(watch_path, {})
103 # if we've already seen an event in this directory (AKA runfolder)
104 # keep track if its already hit the "completed" flag
105 if watch_path_events.has_key(target):
106 run_already_complete = watch_path_events[target].complete
108 watch_path_events[target] = WatcherEvent(target)
109 #self.last_event.setdefault(watch_path, {})[target] = WatcherEvent(target)
111 msg = "Create: %s %s %s %s" % (watch_path, target, event.path, event.name)
113 # the ReadPrep step uses some of the same file completion flags as the
114 # main analysis, which means this completion code might get tripped because of it
115 # so we need to make sure we're getting the completion file in the root of the
117 event_name = event.name.lower()
118 if (event_name in self.completion_files and event.path == runfolder) \
119 or run_already_complete:
120 self.last_event[watch_path][target].complete = True
125 def process_IN_DELETE(self, event):
126 logging.debug("Remove: %s" % os.path.join(event.path, event.name))
129 def process_IN_UNMOUNT(self, event):
130 pathname = os.path.join(event.path, event.name)
131 logging.debug("IN_UNMOUNT: %s" % (pathname,))
132 self.bot.unmount_watch(event.path)
134 class SpoolWatcher(rpc.XmlRpcBot):
136 Watch a directory and send a message when another process is done writing.
138 This monitors a directory tree using inotify (linux specific) and
139 after some files having been written will send a message after <timeout>
140 seconds of no file writing.
142 (Basically when the solexa machine finishes dumping a round of data
143 this'll hopefully send out a message saying hey look theres data available
146 # these params need to be in the config file
147 # I wonder where I should put the documentation
149 # `watchdirs` - list of directories to monitor for modifications
150 # `profile` - specify which .htsworkflow profile to use
151 # `write_timeout` - how many seconds to wait for writes to finish to
153 # `notify_timeout` - how often to timeout from notify
154 # `completion_files` - what files indicates we've finished sequencing
155 # defaults to: netcopy_complete.txt
157 def __init__(self, section=None, configfile=None):
158 #if configfile is None:
159 # self.configfile = "~/.htsworkflow"
160 super(SpoolWatcher, self).__init__(section, configfile)
162 self.cfg['watchdirs'] = None
163 self.cfg['write_timeout'] = 10
164 self.cfg['notify_users'] = None
165 self.cfg['notify_runner'] = None
166 self.cfg['completion_files'] = 'ImageAnalysis_Netcopy_complete_READ2.txt ImageAnalysis_Netcopy_complete_SINGLEREAD.txt'
169 self.watchdir_url_map = {}
170 self.notify_timeout = 0.001
173 self.notify_users = None
174 self.notify_runner = None
177 # keep track if the specified mount point is currently mounted
178 self.mounted_points = {}
179 # keep track of which mount points tie to which watch directories
180 # so maybe we can remount them.
181 self.mounts_to_watches = {}
183 self.eventTasks.append(self.process_notify)
185 def read_config(self, section=None, configfile=None):
186 # Don't give in to the temptation to use logging functions here,
187 # need to wait until after we detach in start
188 super(SpoolWatcher, self).read_config(section, configfile)
190 self.watchdirs = shlex.split(self._check_required_option('watchdirs'))
191 # see if there's an alternate url that should be used for the watchdir
192 for watchdir in self.watchdirs:
193 self.watchdir_url_map[watchdir] = self.cfg.get(watchdir, watchdir)
195 self.write_timeout = int(self.cfg['write_timeout'])
196 self.completion_files = shlex.split(self.cfg['completion_files'])
198 self.notify_users = self._parse_user_list(self.cfg['notify_users'])
200 self.notify_runner = \
201 self._parse_user_list(self.cfg['notify_runner'],
202 require_resource=True)
203 except bot.JIDMissingResource:
204 msg = 'need a full jabber ID + resource for xml-rpc destinations'
205 raise bot.JIDMissingResource(msg)
210 def add_watch(self, watchdirs=None):
212 start watching watchdir or self.watchdir
213 we're currently limited to watching one directory tree.
215 # create the watch managers if we need them
217 self.wm = pyinotify.WatchManager()
218 self.handler = Handler(self.wm, self, self.completion_files)
219 self.notifier = pyinotify.Notifier(self.wm, self.handler)
221 # the one tree limit is mostly because self.wdd is a single item
222 # but managing it as a list might be a bit more annoying
223 if watchdirs is None:
224 watchdirs = self.watchdirs
226 mask = EventsCodes.IN_CREATE | EventsCodes.IN_UNMOUNT
227 # rec traverses the tree and adds all the directories that are there
229 # auto_add will add in new directories as they are created
231 mount_location = mount.find_mount_point_for(w)
232 self.mounted_points[mount_location] = True
233 mounts = self.mounts_to_watches.get(mount_location, [])
236 self.mounts_to_watches[mount_location] = mounts
238 logging.info(u"Watching:"+unicode(w))
239 self.wdds.append(self.wm.add_watch(w, mask, rec=True, auto_add=True))
241 def unmount_watch(self, event_path):
242 # remove backwards so we don't get weirdness from
243 # the list getting shorter
244 for i in range(len(self.wdds),0, -1):
246 logging.info(u'unmounting: '+unicode(wdd.items()))
247 self.wm.rm_watch(wdd.values())
251 def make_copy_url(self, watchdir, list_event_dir):
252 root_copy_url = self.watchdir_url_map[watchdir]
253 if root_copy_url[-1] != '/':
255 copy_url = root_copy_url + list_event_dir
256 logging.debug('Copy url: %s' % (copy_url,))
259 def process_notify(self, *args):
260 if self.notifier is None:
263 # process the queue of events as explained above
264 self.notifier.process_events()
265 #check events waits timeout
266 if self.notifier.check_events(self.notify_timeout):
267 # read notified events and enqeue them
268 self.notifier.read_events()
269 # should we do something?
270 # has something happened?
271 for watchdir, last_events in self.handler.last_event.items():
272 for last_event_dir, last_event_detail in last_events.items():
273 time_delta = time.time() - last_event_detail.time
274 if time_delta > self.write_timeout:
275 print "timeout", unicode(last_event_detail)
276 copy_url = self.make_copy_url(watchdir, last_event_dir)
277 self.startCopy(copy_url)
278 if last_event_detail.complete:
279 self.sequencingFinished(last_event_detail.event_root)
281 self.handler.last_event[watchdir] = {}
282 # handle unmounted filesystems
283 for mount_point, was_mounted in self.mounted_points.items():
284 if not was_mounted and mount.is_mounted(mount_point):
285 # we've been remounted. Huzzah!
287 for watch in self.mounts_to_watches[mount_point]:
288 self.add_watch(watch)
290 "%s was remounted, restarting watch" % \
293 self.mounted_points[mount_point] = True
295 def _parser(self, msg, who):
297 Parse xmpp chat messages
299 help = u"I can send [copy] message, or squencer [finished]"
300 if re.match(u"help", msg):
302 elif re.match("copy", msg):
304 reply = u"sent copy message"
305 elif re.match(u"finished", msg):
308 self.sequencingFinished(words[1])
309 reply = u"sending sequencing finished for %s" % (words[1])
311 reply = u"need runfolder name"
313 reply = u"I didn't understand '%s'" %(msg)
320 # we have to configure pyinotify after BenderJab.start is called
321 # as weird things happen to pyinotify if the stdio is closed
322 # after it's initialized.
324 super(SpoolWatcher, self).run()
330 # destroy the inotify's instance on this interrupt (stop monitoring)
331 if self.notifier is not None:
333 super(SpoolWatcher, self).stop()
335 def startCopy(self, copy_url=None):
336 logging.debug("writes seem to have stopped")
337 if self.notify_runner is not None:
338 for r in self.notify_runner:
339 self.rpc_send(r, tuple([copy_url]), 'startCopy')
340 if self.notify_users is not None:
341 for u in self.notify_users:
342 self.send(u, 'startCopy %s.' % (copy_url,))
344 def sequencingFinished(self, run_dir):
345 # need to strip off self.watchdirs from rundir I suspect.
346 logging.info("run.completed in " + str(run_dir))
347 for watch in self.watchdirs:
348 if not run_dir.startswith(watch):
349 print "%s didn't start with %s" % (run_dir, watch)
351 if watch[-1] != os.path.sep:
353 stripped_run_dir = re.sub(watch, "", run_dir)
355 stripped_run_dir = run_dir
357 logging.debug("stripped to " + stripped_run_dir)
358 if self.notify_users is not None:
359 for u in self.notify_users:
360 self.send(u, 'Sequencing run %s finished' % \
362 if self.notify_runner is not None:
363 for r in self.notify_runner:
364 self.rpc_send(r, (stripped_run_dir,), 'sequencingFinished')
368 return bot.main(args)
370 if __name__ == "__main__":
371 ret = main(sys.argv[1:])
375 # send messages to copier specifying which mount to copy