-import cPickle
-import errno
-import os
+"""A few things that didn't seem to fit anywhere else."""
+
+import os, os.path
import pwd
import tempfile
+import fcntl
+import errno
import threading
+import subprocess
+import shutil
+import sys
+import signal
-from config import PID_FILE
import logger
+PID_FILE = '/var/run/nodemanager.pid'
+
+####################
+def get_default_if():
+ interface = get_if_from_hwaddr(get_hwaddr_from_plnode())
+ if not interface: interface = "eth0"
+ return interface
+
+def get_hwaddr_from_plnode():
+ try:
+ for line in open("/usr/boot/plnode.txt", 'r').readlines():
+ if line.startswith("NET_DEVICE"):
+ return line.split("=")[1].strip().strip('"')
+ except:
+ pass
+ return None
+
+def get_if_from_hwaddr(hwaddr):
+ import sioc
+ devs = sioc.gifconf()
+ for dev in devs:
+ dev_hwaddr = sioc.gifhwaddr(dev)
+ if dev_hwaddr == hwaddr: return dev
+ return None
+####################
+# daemonizing
def as_daemon_thread(run):
+ """Call function <run> with no arguments in its own thread."""
thr = threading.Thread(target=run)
thr.setDaemon(True)
thr.start()
+def close_nonstandard_fds():
+ """Close all open file descriptors other than 0, 1, and 2."""
+ _SC_OPEN_MAX = 4
+ for fd in range(3, os.sysconf(_SC_OPEN_MAX)):
+ try: os.close(fd)
+ except OSError: pass # most likely an fd that isn't open
# after http://www.erlenstar.demon.co.uk/unix/faq_2.html
def daemon():
os.setsid()
if os.fork() != 0: os._exit(0)
os.chdir('/')
- os.umask(0)
+ os.umask(0022)
devnull = os.open(os.devnull, os.O_RDWR)
- for fd in range(3): os.dup2(devnull, fd)
-
-
-def deepcopy(obj):
- """Return a deep copy of obj."""
- return cPickle.loads(cPickle.dumps(obj, -1))
-
+ os.dup2(devnull, 0)
+ # xxx fixme - this is just to make sure that nothing gets stupidly lost - should use devnull
+ crashlog = os.open('/var/log/nodemanager.daemon', os.O_RDWR | os.O_APPEND | os.O_CREAT, 0644)
+ os.dup2(crashlog, 1)
+ os.dup2(crashlog, 2)
def fork_as(su, function, *args):
"""fork(), cd / to avoid keeping unused directories open, close all nonstandard file descriptors (to avoid capturing open sockets), fork() again (to avoid zombies) and call <function> with arguments <args> in the grandchild process. If <su> is not None, set our group and user ids appropriately in the child process."""
if child_pid == 0:
try:
os.chdir('/')
- # close all nonstandard file descriptors
- _SC_OPEN_MAX = 4
- for fd in range(3, os.sysconf(_SC_OPEN_MAX)):
- try: os.close(fd)
- except OSError: pass # most likely an fd that isn't open
- pw_ent = pwd.getpwnam(su)
- os.setegid(pw_ent[3])
- os.seteuid(pw_ent[2])
+ close_nonstandard_fds()
+ if su:
+ pw_ent = pwd.getpwnam(su)
+ os.setegid(pw_ent[3])
+ os.seteuid(pw_ent[2])
child_pid = os.fork()
if child_pid == 0: function(*args)
except:
os.seteuid(os.getuid()) # undo su so we can write the log file
os.setegid(os.getgid())
- logger.log_exc()
+ logger.log_exc("tools: fork_as")
os._exit(0)
else: os.waitpid(child_pid, 0)
-
+####################
+# manage files
def pid_file():
- """We use a pid file to ensure that only one copy of NM is running at a given time. If successful, this function will write a pid file containing the pid of the current process. The return value is the pid of the other running process, or None otherwise."""
+ """We use a pid file to ensure that only one copy of NM is running at a given time.
+If successful, this function will write a pid file containing the pid of the current process.
+The return value is the pid of the other running process, or None otherwise."""
other_pid = None
- # check for a pid file
- if os.access(PID_FILE, os.F_OK):
- # pid file exists, read it
- handle = open(PID_FILE)
+ if os.access(PID_FILE, os.F_OK): # check for a pid file
+ handle = open(PID_FILE) # pid file exists, read it
other_pid = int(handle.read())
handle.close()
# check for a process with that pid by sending signal 0
else: raise # who knows
if other_pid == None:
# write a new pid file
- write_file(PID_FILE, lambda thefile: thefile.write(str(os.getpid())))
+ write_file(PID_FILE, lambda f: f.write(str(os.getpid())))
return other_pid
-
-def write_file(filename, do_write):
+def write_file(filename, do_write, **kw_args):
"""Write file <filename> atomically by opening a temporary file, using <do_write> to write that file, and then renaming the temporary file."""
- os.rename(write_temp_file(do_write), filename)
-
+ shutil.move(write_temp_file(do_write, **kw_args), filename)
-def write_temp_file(do_write):
+def write_temp_file(do_write, mode=None, uidgid=None):
fd, temporary_filename = tempfile.mkstemp()
- thefile = os.fdopen(fd, 'w')
- do_write(thefile)
- thefile.close()
+ if mode: os.chmod(temporary_filename, mode)
+ if uidgid: os.chown(temporary_filename, *uidgid)
+ f = os.fdopen(fd, 'w')
+ try: do_write(f)
+ finally: f.close()
return temporary_filename
+
+# replace a target file with a new contents - checks for changes
+# can handle chmod if requested
+# can also remove resulting file if contents are void, if requested
+# performs atomically:
+# writes in a tmp file, which is then renamed (from sliverauth originally)
+# returns True if a change occurred, or the file is deleted
+def replace_file_with_string (target, new_contents, chmod=None, remove_if_empty=False):
+ try:
+ current=file(target).read()
+ except:
+ current=""
+ if current==new_contents:
+ # if turns out to be an empty string, and remove_if_empty is set,
+ # then make sure to trash the file if it exists
+ if remove_if_empty and not new_contents and os.path.isfile(target):
+ logger.verbose("tools.replace_file_with_string: removing file %s"%target)
+ try: os.unlink(target)
+ finally: return True
+ return False
+ # overwrite target file: create a temp in the same directory
+ path=os.path.dirname(target) or '.'
+ fd, name = tempfile.mkstemp('','repl',path)
+ os.write(fd,new_contents)
+ os.close(fd)
+ if os.path.exists(target):
+ os.unlink(target)
+ shutil.move(name,target)
+ if chmod: os.chmod(target,chmod)
+ return True
+
+
+####################
+# utilities functions to get (cached) information from the node
+
+# get node_id from /etc/planetlab/node_id and cache it
+_node_id=None
+def node_id():
+ global _node_id
+ if _node_id is None:
+ try:
+ _node_id=int(file("/etc/planetlab/node_id").read())
+ except:
+ _node_id=""
+ return _node_id
+
+_root_context_arch=None
+def root_context_arch():
+ global _root_context_arch
+ if not _root_context_arch:
+ sp=subprocess.Popen(["uname","-i"],stdout=subprocess.PIPE)
+ (_root_context_arch,_)=sp.communicate()
+ _root_context_arch=_root_context_arch.strip()
+ return _root_context_arch
+
+
+####################
+class NMLock:
+ def __init__(self, file):
+ logger.log("tools: Lock %s initialized." % file, 2)
+ self.fd = os.open(file, os.O_RDWR|os.O_CREAT, 0600)
+ flags = fcntl.fcntl(self.fd, fcntl.F_GETFD)
+ flags |= fcntl.FD_CLOEXEC
+ fcntl.fcntl(self.fd, fcntl.F_SETFD, flags)
+ def __del__(self):
+ os.close(self.fd)
+ def acquire(self):
+ logger.log("tools: Lock acquired.", 2)
+ fcntl.lockf(self.fd, fcntl.LOCK_SH)
+ def release(self):
+ logger.log("tools: Lock released.", 2)
+ fcntl.lockf(self.fd, fcntl.LOCK_UN)
+
+####################
+# Utilities for getting the IP address of a LXC/Openvswitch slice. Do this by
+# running ifconfig inside of the slice's context.
+
+def get_sliver_process(slice_name, process_cmdline):
+ """ Utility function to find a process inside of an LXC sliver. Returns
+ (cgroup_fn, pid). cgroup_fn is the filename of the cgroup file for
+ the process, for example /proc/2592/cgroup. Pid is the process id of
+ the process. If the process is not found then (None, None) is returned.
+ """
+ try:
+ cmd = 'grep %s /proc/*/cgroup | grep freezer'%slice_name
+ output = os.popen(cmd).readlines()
+ except:
+ # the slice couldn't be found
+ logger.log("get_sliver_process: couldn't find slice %s" % slice_name)
+ return (None, None)
+
+ cgroup_fn = None
+ pid = None
+ for e in output:
+ try:
+ l = e.rstrip()
+ path = l.split(':')[0]
+ comp = l.rsplit(':')[-1]
+ slice_name_check = comp.rsplit('/')[-1]
+
+ if (slice_name_check == slice_name):
+ slice_path = path
+ pid = slice_path.split('/')[2]
+ cmdline = open('/proc/%s/cmdline'%pid).read().rstrip('\n\x00')
+ if (cmdline == process_cmdline):
+ cgroup_fn = slice_path
+ break
+ except:
+ break
+
+ if (not cgroup_fn) or (not pid):
+ logger.log("get_sliver_process: process %s not running in slice %s" % (process_cmdline, slice_name))
+ return (None, None)
+
+ return (cgroup_fn, pid)
+
+def get_sliver_ifconfig(slice_name, device="eth0"):
+ """ return the output of "ifconfig" run from inside the sliver.
+
+ side effects: adds "/usr/sbin" to sys.path
+ """
+
+ # See if setns is installed. If it's not then we're probably not running
+ # LXC.
+ if not os.path.exists("/usr/sbin/setns.so"):
+ return None
+
+ # setns is part of lxcsu and is installed to /usr/sbin
+ if not "/usr/sbin" in sys.path:
+ sys.path.append("/usr/sbin")
+ import setns
+
+ (cgroup_fn, pid) = get_sliver_process(slice_name, "/sbin/init")
+ if (not cgroup_fn) or (not pid):
+ return None
+
+ path = '/proc/%s/ns/net'%pid
+
+ result = None
+ try:
+ setns.chcontext(path)
+
+ args = ["/sbin/ifconfig", device]
+ sub = subprocess.Popen(args, stderr = subprocess.PIPE, stdout = subprocess.PIPE)
+ sub.wait()
+
+ if (sub.returncode != 0):
+ logger.log("get_slice_ifconfig: error in ifconfig: %s" % sub.stderr.read())
+
+ result = sub.stdout.read()
+ finally:
+ setns.chcontext("/proc/1/ns/net")
+
+ return result
+
+def get_sliver_ip(slice_name):
+ ifconfig = get_sliver_ifconfig(slice_name)
+ if not ifconfig:
+ return None
+
+ for line in ifconfig.split("\n"):
+ if "inet addr:" in line:
+ # example: ' inet addr:192.168.122.189 Bcast:192.168.122.255 Mask:255.255.255.0'
+ parts = line.strip().split()
+ if len(parts)>=2 and parts[1].startswith("addr:"):
+ return parts[1].split(":")[1]
+
+ return None
+
+### this returns the kind of virtualization on the node
+# either 'vs' or 'lxc'
+# also caches it in /etc/planetlab/virt for next calls
+# could be promoted to core nm if need be
+virt_stamp="/etc/planetlab/virt"
+def get_node_virt ():
+ try:
+ return file(virt_stamp).read().strip()
+ except:
+ pass
+ logger.log("Computing virt..")
+ try:
+ if subprocess.call ([ 'vserver', '--help' ]) ==0: virt='vs'
+ else: virt='lxc'
+ except:
+ virt='lxc'
+ with file(virt_stamp,"w") as f:
+ f.write(virt)
+ return virt
+
+### this return True or False to indicate that systemctl is present on that box
+# cache result in memory as _has_systemctl
+_has_systemctl=None
+def has_systemctl ():
+ global _has_systemctl
+ if _has_systemctl is None:
+ _has_systemctl = (subprocess.call([ 'systemctl', '--help' ]) == 0)
+ return _has_systemctl
+
+# how to run a command in a slice
+# now this is a painful matter
+# the problem is with capsh that forces a bash command to be injected in its exec'ed command
+# so because lxcsu uses capsh, you cannot exec anything else than bash
+# bottom line is, what actually needs to be called is
+# vs: vserver exec slicename command and its arguments
+# lxc: lxcsu slicename "command and its arguments"
+# which, OK, is no big deal as long as the command is simple enough,
+# but do not stretch it with arguments that have spaces or need quoting as that will become a nightmare
+def command_in_slice (slicename, argv):
+ virt=get_node_virt()
+ if virt=='vs':
+ return [ 'vserver', slicename, 'exec', ] + argv
+ elif virt=='lxc':
+ # wrap up argv in a single string for -c
+ return [ 'lxcsu', slicename, ] + [ " ".join(argv) ]
+ logger.log("command_in_slice: WARNING: could not find a valid virt")
+ return argv
+
+####################
+def init_signals ():
+ def handler (signum, frame):
+ logger.log("Received signal %d - exiting"%signum)
+ os._exit(1)
+ signal.signal(signal.SIGHUP,handler)
+ signal.signal(signal.SIGQUIT,handler)
+ signal.signal(signal.SIGINT,handler)
+ signal.signal(signal.SIGTERM,handler)
git://git.onelab.eu / nodemanager.git / blobdiff