boinc/sched/start

751 lines
23 KiB
Python
Executable File

#!/usr/bin/env python
# -*- mode: python; python-indent: 4; -*-
## $Id$
'''
A program to start/stop BOINC server daemons and run periodic tasks.
Parses config.xml and runs <daemon> and <task> entries.
The format of config.xml is described in boinc/doc/configuration.php.
The main script is "start"; sym-link or hard-link "start" to "stop", "cron".
Invocation methods:
--enable (default if invoked as "start")
Set the project to ENABLED mode and start daemons
--cron If project is in ENABLED mode start daemons and run tasks;
else do nothing.
This command is intended to be run as a real cron job
every five minutes.
--cron-tasks Run tasks but do not start daemons;
This command is intended to be run as a real cron job
every five minutes.
--disable (default if invoked as "stop")
Set project to DISABLED mode and stop daemons.
--status (default if invoked as "status")
Show status.
See "start --help" for options.
Daemons:
These are continuously-running programs.
The process ID is recorded in the <pid_dir> directory
and the process is sent a SIGHUP in a DISABLE operation.
Both tasks and daemons can run on a different host (specified by <host>).
The default is the project's main host, which is specified in config.host
A daemon or task can be turned off by adding the <disabled/> element.
IMPLEMENTATION:
"Start" uses a file "run_state_HOST.xml" that records the enabled mode
and the last run time of periodic tasks on that host.
Looks like this:
<?xml version="1.0">
<boinc>
<tasks>
<task>
<cmd>
get_load
</cmd>
<last_run>
1070392808.86
</last_run>
</task>
...
</tasks>
<enabled>
1
</enabled>
</boinc>
Daemons:
Writes a PID to pid_HOST/command.pid.
Tasks:
Writes a timestamp to run_state_HOST.xml
to remember when the task was last run.
Both:
A lock file (pid_HOST/command.lock)
prevents tasks and daemons from being run
again when they are currently running.
'''
import boinc_path_config
from Boinc import boinc_project_path, configxml
import sys, os, getopt, time, glob, fcntl, signal, socket
right_now = int(time.time())
verbose = os.isatty(sys.stdout.fileno())
verbose_daemon_run = 0
# how long (in seconds) parent should wait before continuing after a fork.
# this is just a safety measure in case anything doesn't play nice if
# starting simultaneously. also it keeps output in sequence.
fork_delay = 0.1
ignore_timestamps = False
prune_run_state = True #
is_main_host = False
local_hostname = ''
delegate_other_hosts_in_parallel = False
def get_host_list():
'''
Get a list of hosts mentioned in <host> elements of tasks and daemons
'''
hosts = []
for task in config.tasks:
host = task.__dict__.get('host')
if (host and (host not in hosts)):
hosts.append(host)
for task in config.daemons:
host = task.__dict__.get('host')
if (host and (host not in hosts)):
hosts.append(host)
return hosts
def assign_task_defaults():
'''
Set the "host" and "disabled" attribute of tasks and daemons
'''
for task in config.tasks:
host = task.__dict__.get('host')
if not host:
task.host = config.config.host
disabled = task.__dict__.get('disabled')
if disabled and disabled != "0":
task.disabled = 1
else:
task.disabled = 0
always_run = task.__dict__.get('always_run')
if always_run and always_run != "0":
task.always_run = 1
else:
task.always_run = 0
for task in config.daemons:
host = task.__dict__.get('host')
if not host:
task.host = config.config.host
disabled = task.__dict__.get('disabled')
if disabled and disabled != "0":
task.disabled = 1
else:
task.disabled = 0
def get_dir(name):
return config.config.__dict__.get(name+'_dir') or boinc_project_path.project_path(name)
def ensure_get_dir(name):
f = get_dir(name)
ensure_dir(f)
return f
def is_daemon(task):
'''returns true if task is a daemon'''
return task._name == 'daemon'
def get_task_command_basename(task):
return os.path.basename(task.cmd.split()[0])
def get_task_output_name(task):
return os.path.join(log_dir,
task.__dict__.get('output') or get_task_command_basename(task) + '.out')
def get_daemon_output_name(task):
return os.path.join(log_dir,
task.__dict__.get('output') or get_task_command_basename(task) + '.log')
def get_daemon_pid_name(task):
return os.path.join(pid_dir,
task.__dict__.get('pid_file') or get_task_command_basename(task) + '.pid')
def output_is_file(filename):
return filename and not filename.startswith('/dev/')
def get_task_lock_name(task):
return os.path.join(pid_dir,
task.__dict__.get('lock_file')
or (output_is_file(task.__dict__.get('output')) and task.__dict__.get('output')+'.lock')
or get_task_command_basename(task) + '.lock')
def ensure_dir(filename):
try:
os.mkdir(filename)
except OSError:
return
def timestamp(t = None):
return time.strftime('%Y/%m/%d %H:%M:%S', time.localtime(t or right_now))
def safe_read_int(filename):
try:
return int(open(filename).readline().strip())
except:
return 0
def get_stop_daemons_filename():
return boinc_project_path.project_path('stop_daemons')
def get_stop_sched_filename():
return boinc_project_path.project_path('stop_sched')
def write_stop_daemons():
print >>open(get_stop_daemons_filename(),'w'), '<stop/>'
def remove_stop_daemons():
if os.path.exists(get_stop_daemons_filename()):
os.unlink(get_stop_daemons_filename())
def write_stop_sched():
print >>open(get_stop_sched_filename(),'w'), '<stop/>'
def remove_stop_sched():
if os.path.exists(get_stop_sched_filename()):
os.unlink(get_stop_sched_filename())
def safe_unlink(filename):
try:
os.unlink(filename)
except OSError, e:
print "Couldn't unlink %s:"%filename,e
def remove_cached_home_page():
path = boinc_project_path.project_path('html/cache/65/index.php')
if os.path.exists(path):
print 'removing ' + path
safe_unlink(path)
def redirect(stdout='/dev/null', stderr=None, stdin='/dev/null'):
'''
Redirects stdio. The stdin, stdout, and stderr arguments are file names
that will be opened and be used to replace the standard file descriptors
in sys.stdin, sys.stdout, and sys.stderr. These arguments are optional
and default to /dev/null.
'''
si = open(stdin, 'r')
if not stderr: stderr = stdout
se = open(stderr, 'a+', 0)
if stdout == stderr:
so = se
else:
so = open(stdout, 'a+')
# Redirect standard file descriptors.
os.dup2(si.fileno(), sys.stdin.fileno())
os.dup2(so.fileno(), sys.stdout.fileno())
os.dup2(se.fileno(), sys.stderr.fileno())
def fork():
''' fork with fork_delay '''
pid = os.fork()
if pid:
time.sleep(fork_delay)
return pid
def double_fork():
'''
This forks the current process into a daemon using a double-fork.
Returns 1 for parent, 0 for child.
See: http://www.erlenstar.demon.co.uk/unix/faq_2.html#SEC16
http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/66012
'''
# this is necessary because otherwise any buffered output would get
# printed twice after the fork!
sys.stdout.flush()
# Do first fork.
try:
pid = fork()
if pid > 0: return 1
except OSError, e:
sys.stderr.write("fork #1 failed: (%d) %s\n" % (e.errno, e.strerror))
sys.exit(1)
# Decouple from parent environment.
os.chdir("/")
os.umask(0)
os.setsid()
# Do second fork.
try:
pid = os.fork()
if pid > 0: os._exit(0) # Exit second parent.
except OSError, e:
sys.stderr.write("fork #2 failed: (%d) %s\n" % (e.errno, e.strerror))
os._exit(1)
return 0
def write_pid_file(pidfile):
print >>open(pidfile,'w'), os.getpid()
def is_pid_running(pid):
try:
os.kill(pid,0)
return True
except OSError:
return False
# if we ever want to use this on windows see:
# http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/65203
# returns 0 on success, -1 on error
locks = []
def lock_file(filename):
global locks
file = open(filename,'w')
locks.append(file)
try:
return fcntl.lockf(file.fileno(), fcntl.LOCK_EX|fcntl.LOCK_NB)
except IOError:
return -1
def is_lock_file_locked(filename):
if lock_file(filename):
return True
else:
os.unlink(filename)
def contains_shell_characters(command):
return ('"' in command or "'" in command or
'\\' in command or '|' in command or
'>' in command)
def exec_command_string(command):
args = command.strip().split()
os.chdir(tmp_dir)
try:
if contains_shell_characters(command):
os.execl('/bin/sh', 'sh', '-c', ' '.join(args))
else:
os.execvp( args[0], args )
# on success we don't reach here
print >>sys.stderr, "Couldn't exec '%s'"%command
except OSError, e:
print >>sys.stderr, "Couldn't execute '%s':" %command, e
os._exit(1)
def lookup_task_run_state(task):
for run_state_task in run_state.tasks:
if run_state_task.cmd == task.cmd:
run_state_task.prunable = False
return run_state_task
run_state_task = run_state.tasks.make_node_and_append('task')
run_state_task.cmd = task.cmd
run_state_task.last_run = 0
run_state_task.prunable = False
return run_state_task
def interpret_period(str):
''' "5 min" -> 5*60 ; "1 hour" -> 1*60*60; "2" -> 2*60 '''
s = str.strip().split()
try:
num = int(s[0])
if len(s) == 1:
return num*60
if len(s) == 2:
u = s[1].lower()
if u.startswith('s'):
return num
if u.startswith('m'):
return num*60
if u.startswith('h'):
return num*60*60
if u.startswith('d'):
return num*60*60*24
if u.startswith('w'):
return num*60*60*24*7
if u.startswith('mo'):
return num*60*60*24*30
except ValueError:
pass
raise SystemExit('Invalid task period "%s"'%str)
def when_will_task_next_run(task, task_run_state):
return float(task_run_state.last_run) + interpret_period(task.period)
def time_to_run_task(task, task_run_state):
return (ignore_timestamps
or (right_now >= when_will_task_next_run(task,task_run_state)))
def update_task_timestamp(task_run_state):
task_run_state.last_run = right_now
def run_task(task):
'''Fork and exec command without stdout/err redirection'''
task_run_state = lookup_task_run_state(task)
if not time_to_run_task(task, task_run_state):
if verbose:
print " Not running task because not time yet:",task.cmd
return
if verbose:
print " Running task:", task.cmd
update_task_timestamp(task_run_state)
# we don't need the full double-fork because this should finish quickly
if fork() > 0: return
if lock_file(get_task_lock_name(task)):
print >>sys.stderr, " Task currently running! (%s)"%task.cmd
sys.exit(1)
redirect(get_task_output_name(task))
exec_command_string(task.cmd)
def run_daemon(task):
'''Double-fork and exec command with stdout/err redirection and pid writing'''
if double_fork() > 0: return
if lock_file(get_task_lock_name(task)):
if verbose:
print >>sys.stderr, " Daemon already running:",task.cmd
sys.exit(0)
if verbose or verbose_daemon_run:
print " Starting daemon:", task.cmd
sys.stdout.flush()
redirect(get_daemon_output_name(task))
write_pid_file(get_daemon_pid_name(task))
print "[%s] Executing command:"%timestamp(), task.cmd
sys.stdout.flush()
exec_command_string(task.cmd)
def run_daemons():
found_any = False
if verbose: print "Starting daemons"
if is_main_host:
remove_stop_daemons()
for task in config.daemons:
if task.host != local_hostname:
continue
found_any = True
if task.disabled:
continue
run_daemon(task)
return found_any
def run_tasks():
if verbose: print "Running tasks"
prepare_run_state_pruning()
for task in config.tasks:
if task.host != local_hostname:
continue
if task.disabled:
continue
if task.always_run:
run_task(task)
continue
if run_state.enabled:
run_task(task)
do_prune_run_states()
def prepare_run_state_pruning():
for run_state_task in run_state.tasks:
run_state_task.prunable = True
def do_prune_run_states():
'''\
Delete tasks that have prunable==True (since we didn't touch them this run)
'''
if not prune_run_state:
return
for run_state_task in run_state.tasks:
if run_state_task.prunable:
# print 'Deleting obsolete run_state task', run_state_task.cmd, '(last run %s)' %timestamp(float(run_state_task.last_run))
run_state.tasks.remove_node(run_state_task)
def stop_daemon(pid):
'''returns 1 if something stopped, else 0'''
try:
os.kill(pid, signal.SIGHUP)
except OSError, e:
if e.errno != 3:
print >>sys.stderr, "Warning: couldn't kill pid %d:"%pid, e
return 0
if verbose:
print " Killed process", pid
return 1
def wait_for_process_to_end(pid):
'''if process is still running, wait for it to end.'''
if is_pid_running(pid):
if verbose:
print " Waiting for process",pid,"to end: ",
sys.stdout.flush()
while is_pid_running(pid):
if verbose:
sys.stdout.write('.')
sys.stdout.flush()
time.sleep(.5)
if verbose:
print ' ok'
def stop_daemons():
if verbose: print "Stopping all daemons"
write_stop_daemons()
pid_files = glob.glob(os.path.join(pid_dir, '*.pid'))
count = 0
pids = map(safe_read_int, pid_files)
map(safe_unlink, pid_files)
for pid in pids:
count += stop_daemon(pid)
for pid in pids:
wait_for_process_to_end(pid)
if verbose:
if not count:
print " (No processes stopped)"
######################################################################
## command (action) functions:
def command_enable_start():
if verbose:
if run_state.enabled:
print "Staying in ENABLED mode"
else:
print "Entering ENABLED mode"
# NOTE: must use '1', not '0' here
run_state.enabled = 1
found_any = run_daemons()
if not found_any:
print "No daemons for this host found - check host name in config.xml"
time.sleep(10) # give feeder time to start up and create shmem
remove_stop_sched()
remove_cached_home_page()
def command_cron_start():
if verbose: print "Verbose cron-start: status ==", (run_state.enabled and 'ENABLED' or 'DISABLED')
global verbose_daemon_run
if run_state.enabled:
verbose_daemon_run = 1
run_daemons()
run_tasks()
else:
verbose_daemon_run = 1
run_tasks()
def command_tasks_start():
if verbose: print "Verbose cron-start: status ==", (run_state.enabled and 'ENABLED' or 'DISABLED')
run_tasks()
def command_disable_stop():
if verbose:
if run_state.enabled:
print "Entering DISABLED mode"
else:
print "Staying in DISABLED mode"
run_state.enabled = 0
write_stop_sched() # do this before stop_daemons(),
# else scheduler will complained about no shared mem
stop_daemons()
remove_cached_home_page()
def command_status():
if run_state.enabled:
print "BOINC is ENABLED"
else:
print "BOINC is DISABLED"
if verbose:
print
print "DAEMON pid status lockfile disabled commandline"
n = 0
for task in config.daemons:
if task.host != local_hostname:
continue
n += 1
pid = safe_read_int(get_daemon_pid_name(task)) or 0
if not pid:
rs = " "
elif is_pid_running(pid):
rs = " running "
else:
rs = "NOT RUNNING"
if is_lock_file_locked(get_task_lock_name(task)):
lu = " locked "
else:
lu = "UNLOCKED"
if task.disabled:
d = " yes "
else:
d = " no "
print " %2d"%n, " %5d"%pid, rs, lu, d, " ", task.cmd
print
print "TASK last run period next run lock file disabled commandline"
n = 0
for task in config.tasks:
if task.host != local_hostname:
continue
n += 1
task_run_state = lookup_task_run_state(task)
when_last_run = float(task_run_state.last_run)
last_run = when_last_run and timestamp(when_last_run) or '?'
when_next_run = when_will_task_next_run(task, lookup_task_run_state(task))
next_run = (when_next_run <= right_now) and 'NOW' or timestamp(when_next_run)
if is_lock_file_locked(get_task_lock_name(task)):
lu = " LOCKED "
else:
lu = "unlocked"
if task.disabled:
d = " yes "
else:
d = " no "
print " %2d"%n, last_run.center(20), task.period.ljust(10), \
next_run.center(20), lu, d, " ", task.cmd
pass
def command_show_config():
# TODO: - all config items (e.g. where's logdir?)
raise SystemExit('TODO')
# ------------- main program begins here ---------------------
local_hostname = socket.gethostname()
local_hostname = local_hostname.split('.')[0]
# print 'local hostname: ', local_hostname
cwd = os.getcwd()
program_name = os.path.basename(sys.argv[0])
if program_name == 'start':
command = command_enable_start
elif program_name == 'stop':
command = command_disable_stop
elif program_name == 'status':
command = command_status
else:
command = None
def help():
print >>sys.stderr, "Syntax: %s [options] [command]" % sys.argv[0]
print >>sys.stderr, """ Starts or stops BOINC daemons and tasks.
Commands:
--enable (-e) Set BOINC to ENABLED mode and start daemons
--cron (-c) If ENABLED, start daemons and run tasks
Intended to be run from real cron every 5 min.
--cron-tasks If ENABLED run tasks only and do not start daemons
Intended to be run from real cron every 5 min.
--disable (-d) Set BOINC to DISABLED mode and stop daemons
--status (-s) Show status.
--show-config Show configuration
Options:
--quiet (-q) Operate quietly, even if STDOUT is a tty.
--verbose (-v) Operate verbosely, even if STDOUT is not a tty.
--config-file= Use specified file instead of program-path/../config.xml
--run-state-file= Use specified file instead of program-path/../run_state.xml
--fork-delay= Seconds to sleep between daemon forks instead of 0.1
--ignore-timestamps Ignore timestamps; for cron mode, runs all tasks now
--prune-run-state Delete unused timestamps in run_state.xml
"""
if program_name == 'start':
print >>sys.stderr, "Based on the invocation name as `start', the default action is --enable."
elif program_name == 'stop':
print >>sys.stderr, "Based on the invocation name as `stop', the default action is --disable."
elif program_name == 'status':
print >>sys.stderr, "Based on the invocation name as `status', the default action is --status."
sys.exit(1)
config_filename = boinc_project_path.config_xml_filename
run_state_filename = boinc_project_path.run_state_xml_filename
try:
opts, args = getopt.getopt(sys.argv[1:], 'cedskqvh?',
('enable', 'cron', 'cron-tasks', 'disable',
'start', 'stop', 'kill', 'status',
'show-config',
'ignore-timestamps',
'fork-delay=',
'config-file=', 'run-state-file=',
'prune-run-state',
'quiet', 'verbose', 'help'))
except Exception, e:
print >>sys.stderr, e
print >>sys.stderr, "Use '%s --help' for help" % sys.argv[0]
sys.exit(1)
for opt,v in opts:
if opt == '-q' or opt == '--quiet':
verbose = 0
elif opt == '-v' or opt == '--verbose':
verbose = 1
elif opt == '-h' or opt == '--help' or opt == '-?':
help()
elif opt == '-e' or opt == '--enable' or opt == '--start':
command = command_enable_start
elif opt == '-c' or opt == '--cron':
command = command_cron_start
elif opt == '--cron-task':
command = command_tasks_start
elif opt == '-d' or opt == '--disable' or opt == '--stop' or opt == '-k' or opt == '--kill':
command = command_disable_stop
elif opt == '-s' or opt == '--status':
command = command_status
elif opt == '--show-config':
command = command_show_config
elif opt == '--ignore-timestamps':
ignore_timestamps = True
elif opt == '--prune-run-state':
prune_run_state = True
elif opt == '--config-file':
config_filename = v
elif opt == '--run-state-file':
run_state_filename = v
elif opt == '--fork-delay':
fork_delay = v
else: assert(False)
if not command:
raise SystemExit('No command specified and script name is not "start", "stop", or "status"')
config = configxml.ConfigFile(config_filename).read()
run_state = configxml.RunStateFile(run_state_filename).read(failopen_ok = True)
os.chdir(boinc_project_path.project_path())
bin_dir = get_dir('bin')
cgi_bin_dir = get_dir('cgi_bin')
tmp_dir = ensure_get_dir('tmp_'+local_hostname)
log_dir = ensure_get_dir('log_'+local_hostname)
pid_dir = ensure_get_dir('pid_'+local_hostname)
is_main_host = config.config.host == local_hostname
if 'PATH' in os.environ:
os.environ['PATH'] = bin_dir + ':' + os.environ['PATH']
else:
os.environ['PATH'] = bin_dir
start_lockfile = os.path.join(pid_dir, 'start.lock.'+local_hostname)
if lock_file(start_lockfile):
print >>sys.stderr, "start is currently running!"
sys.exit(1)
assign_task_defaults()
apply(command)
run_state.write()
if is_main_host:
if delegate_other_hosts_in_parallel:
wait_mode = os.P_NOWAIT
else:
wait_mode = os.P_WAIT
other_hosts = get_host_list()
for host in other_hosts:
if host == local_hostname:
continue
remote_cmd = [ 'ssh', host, 'cd', cwd, ' && ' ] + sys.argv
if verbose:
remote_cmd += [ '-v' ]
print 'running ', ' '.join(remote_cmd)
os.spawnvp(wait_mode, remote_cmd[0], remote_cmd)
os.unlink(start_lockfile)