import logging
import os
import re
+import shlex
import sys
import time
-#import glob
from htsworkflow.util import mount
from benderjab import rpc
+def get_top_dir(root, path):
+ """
+ Return the directory in path that is a subdirectory of root.
+ e.g.
+
+ >>> print get_top_dir('/a/b/c', '/a/b/c/d/e/f')
+ d
+ >>> print get_top_dir('/a/b/c/', '/a/b/c/d/e/f')
+ d
+ >>> print get_top_dir('/a/b/c', '/g/e/f')
+ None
+ >>> print get_top_dir('/a/b/c', '/a/b/c')
+ <BLANKLINE>
+ """
+ if path.startswith(root):
+ subpath = path[len(root):]
+ if subpath.startswith('/'):
+ subpath = subpath[1:]
+ return subpath.split(os.path.sep)[0]
+ else:
+ return None
class WatcherEvents(object):
# two events need to be tracked
class Handler(pyinotify.ProcessEvent):
- def __init__(self, watchmanager, bot):
- self.last_event_time = None
+ def __init__(self, watchmanager, bot, ipar=False):
+ """
+ ipar flag indicates we should wait for ipar to finish, instead of
+ just the run finishing
+ """
+ self.last_event = {}
self.watchmanager = watchmanager
self.bot = bot
+ self.ipar_mode = ipar
+ if self.ipar_mode:
+ self.last_file = 'IPAR_Netcopy_Complete.txt'.lower()
+ else:
+ self.last_file = "run.completed".lower()
def process_IN_CREATE(self, event):
- self.last_event_time = time.time()
- msg = "Create: %s" % os.path.join(event.path, event.name)
- if event.name.lower() == "run.completed":
- try:
- self.bot.sequencingFinished(event.path)
- except IOError, e:
- logging.error("Couldn't send sequencingFinished")
- logging.debug(msg)
+ for wdd in self.bot.wdds:
+ for watch_path in self.bot.watchdirs:
+ if event.path.startswith(watch_path):
+ target = get_top_dir(watch_path, event.path)
+ self.last_event.setdefault(watch_path, {})[target] = time.time()
+
+ msg = "Create: %s %s" % (event.path, event.name)
+
+ if event.name.lower() == self.last_file:
+ try:
+ self.bot.sequencingFinished(event.path)
+ except IOError, e:
+ pass
+ logging.error("Couldn't send sequencingFinished")
+ logging.debug(msg)
def process_IN_DELETE(self, event):
logging.debug("Remove: %s" % os.path.join(event.path, event.name))
+ pass
def process_IN_UNMOUNT(self, event):
pathname = os.path.join(event.path, event.name)
logging.debug("IN_UNMOUNT: %s" % (pathname,))
- self.bot.unmount_watch()
+ self.bot.unmount_watch(event.path)
class SpoolWatcher(rpc.XmlRpcBot):
"""
# these params need to be in the config file
# I wonder where I should put the documentation
#:Parameters:
- # `watchdir` - which directory tree to monitor for modifications
+ # `watchdirs` - list of directories to monitor for modifications
# `profile` - specify which .htsworkflow profile to use
# `write_timeout` - how many seconds to wait for writes to finish to
# the spool
# self.configfile = "~/.htsworkflow"
super(SpoolWatcher, self).__init__(section, configfile)
- self.cfg['watchdir'] = None
+ self.cfg['watchdirs'] = None
self.cfg['write_timeout'] = 10
self.cfg['notify_users'] = None
self.cfg['notify_runner'] = None
-
+ self.cfg['wait_for_ipar'] = 0
+
+ self.watchdirs = []
+ self.watchdir_url_map = {}
self.notify_timeout = 0.001
- self.wm = pyinotify.WatchManager()
- self.handler = Handler(self.wm, self)
- self.notifier = pyinotify.Notifier(self.wm, self.handler)
- self.wdd = None
- self.mount_point = None
- self.mounted = True
-
+
+ self.wm = None
self.notify_users = None
self.notify_runner = None
+ self.wdds = []
+
+ # keep track if the specified mount point is currently mounted
+ self.mounted_points = {}
+ # keep track of which mount points tie to which watch directories
+ # so maybe we can remount them.
+ self.mounts_to_watches = {}
self.eventTasks.append(self.process_notify)
def read_config(self, section=None, configfile=None):
super(SpoolWatcher, self).read_config(section, configfile)
- self.watch_dir = self._check_required_option('watchdir')
+ self.watchdirs = shlex.split(self._check_required_option('watchdirs'))
+ # see if there's an alternate url that should be used for the watchdir
+ for watchdir in self.watchdirs:
+ self.watchdir_url_map[watchdir] = self.cfg.get(watchdir, watchdir)
+
self.write_timeout = int(self.cfg['write_timeout'])
+ self.wait_for_ipar = int(self.cfg['wait_for_ipar'])
self.notify_users = self._parse_user_list(self.cfg['notify_users'])
try:
require_resource=True)
except bot.JIDMissingResource:
msg = 'need a full jabber ID + resource for xml-rpc destinations'
- logging.FATAL(msg)
raise bot.JIDMissingResource(msg)
- def add_watch(self, watchdir=None):
+ self.handler = None
+ self.notifier = None
+
+ def add_watch(self, watchdirs=None):
"""
- start watching watchdir or self.watch_dir
+ start watching watchdir or self.watchdir
we're currently limited to watching one directory tree.
"""
+ # create the watch managers if we need them
+ if self.wm is None:
+ self.wm = pyinotify.WatchManager()
+ self.handler = Handler(self.wm, self, self.wait_for_ipar)
+ self.notifier = pyinotify.Notifier(self.wm, self.handler)
+
# the one tree limit is mostly because self.wdd is a single item
# but managing it as a list might be a bit more annoying
- if watchdir is None:
- watchdir = self.watch_dir
- logging.info("Watching:"+str(watchdir))
-
- self.mount_point = mount.find_mount_point_for(watchdir)
+ if watchdirs is None:
+ watchdirs = self.watchdirs
mask = EventsCodes.IN_CREATE | EventsCodes.IN_UNMOUNT
# rec traverses the tree and adds all the directories that are there
# at the start.
# auto_add will add in new directories as they are created
- self.wdd = self.wm.add_watch(watchdir, mask, rec=True, auto_add=True)
+ for w in watchdirs:
+ mount_location = mount.find_mount_point_for(w)
+ self.mounted_points[mount_location] = True
+ mounts = self.mounts_to_watches.get(mount_location, [])
+ if w not in mounts:
+ mounts.append(w)
+ self.mounts_to_watches[mount_location] = mounts
+
+ logging.info(u"Watching:"+unicode(w))
+ self.wdds.append(self.wm.add_watch(w, mask, rec=True, auto_add=True))
- def unmount_watch(self):
- if self.wdd is not None:
- self.wm.rm_watch(self.wdd.values())
- self.wdd = None
- self.mounted = False
+ def unmount_watch(self, event_path):
+ # remove backwards so we don't get weirdness from
+ # the list getting shorter
+ for i in range(len(self.wdds),0, -1):
+ wdd = self.wdds[i]
+ logging.info(u'unmounting: '+unicode(wdd.items()))
+ self.wm.rm_watch(wdd.values())
+ del self.wdds[i]
+ self.mounted = False
def process_notify(self, *args):
+ if self.notifier is None:
+ # nothing to do yet
+ return
# process the queue of events as explained above
self.notifier.process_events()
#check events waits timeout
self.notifier.read_events()
# should we do something?
# has something happened?
- last_event_time = self.handler.last_event_time
- if last_event_time is not None:
- time_delta = time.time() - last_event_time
- if time_delta > self.write_timeout:
- self.startCopy()
- self.handler.last_event_time = None
+ for watchdir, last_events in self.handler.last_event.items():
+ #logging.debug('last_events: %s %s' % (watchdir, last_events))
+ for last_event_dir, last_event_time in last_events.items():
+ time_delta = time.time() - last_event_time
+ if time_delta > self.write_timeout:
+ self.startCopy(watchdir, last_event_dir)
+ self.handler.last_event[watchdir] = {}
# handle unmounted filesystems
- if not self.mounted:
- if mount.is_mounted(self.mount_point):
+ for mount_point, was_mounted in self.mounted_points.items():
+ if not was_mounted and mount.is_mounted(mount_point):
# we've been remounted. Huzzah!
# restart the watch
- self.add_watch()
- self.mounted = True
- logging.info(
- "%s was remounted, restarting watch" % \
- (self.mount_point)
- )
+ for watch in self.mounts_to_watches[mount_point]:
+ self.add_watch(watch)
+ logging.info(
+ "%s was remounted, restarting watch" % \
+ (mount_point)
+ )
+ self.mounted_points[mount_point] = True
def _parser(self, msg, who):
"""
reply = u"I didn't understand '%s'" %(msg)
return reply
- def start(self, daemonize):
+ def run(self):
"""
Start application
"""
+ # we have to configure pyinotify after BenderJab.start is called
+ # as weird things happen to pyinotify if the stdio is closed
+ # after it's initialized.
self.add_watch()
- super(SpoolWatcher, self).start(daemonize)
+ super(SpoolWatcher, self).run()
def stop(self):
"""
shutdown application
"""
# destroy the inotify's instance on this interrupt (stop monitoring)
- self.notifier.stop()
+ if self.notifier is not None:
+ self.notifier.stop()
super(SpoolWatcher, self).stop()
- def startCopy(self):
+ def startCopy(self, watchdir=None, event_path=None):
logging.debug("writes seem to have stopped")
if self.notify_runner is not None:
for r in self.notify_runner:
self.rpc_send(r, tuple(), 'startCopy')
+ if self.notify_users is not None:
+ for u in self.notify_users:
+ self.send(u, 'startCopy %s %s' % (watchdir, event_path))
def sequencingFinished(self, run_dir):
- # need to strip off self.watch_dir from rundir I suspect.
+ # need to strip off self.watchdirs from rundir I suspect.
logging.info("run.completed in " + str(run_dir))
pattern = self.watch_dir
if pattern[-1] != os.path.sep:
return bot.main(args)
if __name__ == "__main__":
- sys.exit(main(sys.argv[1:]))
+ ret = main(sys.argv[1:])
+ #sys.exit(ret)
+# TODO:
+# send messages to copier specifying which mount to copy