X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=tools.py;h=d8f216e838b2f82a96f24326f535832f9d87b9f8;hb=74d5761f260b5d887733e826b4ce6aed11f11487;hp=e88a1b2bcda23269fc4e1f2eec0aef471b18b19b;hpb=8ab3434424c269cc2ec7cc47f7705099c9462117;p=nodemanager.git diff --git a/tools.py b/tools.py index e88a1b2..d8f216e 100644 --- a/tools.py +++ b/tools.py @@ -2,7 +2,8 @@ """A few things that didn't seem to fit anywhere else.""" -import os, os.path +import os +import os.path import pwd import tempfile import fcntl @@ -18,11 +19,15 @@ 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" + if not interface: + interface = "eth0" return interface + def get_hwaddr_from_plnode(): try: for line in open("/usr/boot/plnode.txt", 'r').readlines(): @@ -32,44 +37,57 @@ def get_hwaddr_from_plnode(): 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 + if dev_hwaddr == hwaddr: + return dev return None #################### # daemonizing + + def as_daemon_thread(run): """Call function 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 + 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(): """Daemonize the current process.""" - if os.fork() != 0: os._exit(0) + if os.fork() != 0: + os._exit(0) os.setsid() - if os.fork() != 0: os._exit(0) + if os.fork() != 0: + os._exit(0) os.chdir('/') - os.umask(0022) + os.umask(0o022) devnull = os.open(os.devnull, os.O_RDWR) 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) + crashlog = os.open('/var/log/nodemanager.daemon', + os.O_RDWR | os.O_APPEND | os.O_CREAT, 0o644) os.dup2(crashlog, 1) os.dup2(crashlog, 2) + def fork_as(su, function, *args): """ fork(), cd / to avoid keeping unused directories open, @@ -89,16 +107,20 @@ If is not None, set our group and user ids os.setegid(pw_ent[3]) os.seteuid(pw_ent[2]) child_pid = os.fork() - if child_pid == 0: function(*args) + 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("tools: fork_as") os._exit(0) - else: os.waitpid(child_pid, 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. @@ -111,15 +133,19 @@ The return value is the pid of the other running process, or None otherwise. other_pid = int(handle.read()) handle.close() # check for a process with that pid by sending signal 0 - try: os.kill(other_pid, 0) - except OSError, e: - if e.errno == errno.ESRCH: other_pid = None # doesn't exist - else: raise # who knows + try: + os.kill(other_pid, 0) + except OSError as e: + if e.errno == errno.ESRCH: + other_pid = None # doesn't exist + else: + raise # who knows if other_pid == None: # write a new pid file write_file(PID_FILE, lambda f: f.write(str(os.getpid()))) return other_pid + def write_file(filename, do_write, **kw_args): """ Write file atomically by opening a temporary file, @@ -127,16 +153,23 @@ using to write that file, and then renaming the temporary file. """ shutil.move(write_temp_file(do_write, **kw_args), filename) + def write_temp_file(do_write, mode=None, uidgid=None): fd, temporary_filename = tempfile.mkstemp() - if mode: os.chmod(temporary_filename, mode) - if uidgid: os.chown(temporary_filename, *uidgid) + 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() + try: + do_write(f) + finally: + f.close() return temporary_filename -def replace_file_with_string (target, new_contents, chmod=None, remove_if_empty=False): + +def replace_file_with_string(target, new_contents, + chmod=None, remove_if_empty=False): """ Replace a target file with a new contents checks for changes: does not do anything if previous state was already right @@ -147,65 +180,80 @@ writes in a tmp file, which is then renamed (from sliverauth originally) returns True if a change occurred, or the file is deleted """ try: - current=file(target).read() + with open(target) as feed: + current = feed.read() except: - current="" - if current==new_contents: + 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 + logger.verbose( + "tools.replace_file_with_string: removing file {}".format(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) + path = os.path.dirname(target) or '.' + fd, name = tempfile.mkstemp('', 'repl', path) + os.write(fd, new_contents.encode()) os.close(fd) if os.path.exists(target): os.unlink(target) - shutil.move(name,target) - if chmod: os.chmod(target,chmod) + 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 +_node_id = None + + def node_id(): global _node_id if _node_id is None: try: - _node_id=int(file("/etc/planetlab/node_id").read()) + with open("/etc/planetlab/node_id") as f: + _node_id = int(f.read()) except: - _node_id="" + _node_id = "" return _node_id -_root_context_arch=None + +_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() + 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) + logger.log("tools: Lock {} initialized.".format(file), 2) + self.fd = os.open(file, os.O_RDWR | os.O_CREAT, 0o600) 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) @@ -214,18 +262,21 @@ class NMLock: # 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. + """ + 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 + cmd = 'grep {} /proc/*/cgroup | grep freezer'.format(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) + logger.log( + "get_sliver_process: couldn't find slice {}".format(slice_name)) return (None, None) cgroup_fn = None @@ -237,15 +288,30 @@ def get_sliver_process(slice_name, process_cmdline): comp = l.rsplit(':')[-1] slice_name_check = comp.rsplit('/')[-1] # the lines below were added by Guilherme - # due to the ipv6 plugin requirements (LXC) - virt=get_node_virt() - if virt=='lxc': - slice_name_check = slice_name_check.rsplit('.')[0] + # due to the LXC requirements + # What we have to consider here is that libervirt on Fedora 18 + # uses the following line: + # /proc/1253/cgroup:6:freezer:/machine.slice/auto_sirius.libvirt-lxc + # While the libvirt on Fedora 20 and 21 uses the following line: + # /proc/1253/cgroup:6:freezer:/machine.slice/machine-lxc\x2del_sirius.scope + # Further documentation on: + # https://libvirt.org/cgroups.html#systemdScope + virt = get_node_virt() + if virt == 'lxc': + # This is for Fedora 20 or later + regexf20orlater = re.compile(r'machine-lxc\\x2d(.+).scope') + isf20orlater = regexf20orlater.search(slice_name_check) + if isf20orlater: + slice_name_check = isf20orlater.group(1) + else: + # This is for Fedora 18 + slice_name_check = slice_name_check.rsplit('.')[0] if (slice_name_check == slice_name): slice_path = path pid = slice_path.split('/')[2] - cmdline = open('/proc/%s/cmdline'%pid).read().rstrip('\n\x00') + with open('/proc/{}/cmdline'.format(pid)) as cmdfile: + cmdline = cmdfile.read().rstrip('\n\x00') if (cmdline == process_cmdline): cgroup_fn = slice_path break @@ -253,7 +319,8 @@ def get_sliver_process(slice_name, process_cmdline): break if (not cgroup_fn) or (not pid): - logger.log("get_sliver_process: process %s not running in slice %s" % (process_cmdline, slice_name)) + logger.log("get_sliver_process: process {} not running in slice {}" + .format(process_cmdline, slice_name)) return (None, None) return (cgroup_fn, pid) @@ -262,6 +329,7 @@ def get_sliver_process(slice_name, process_cmdline): # Added by Guilherme Sperb Machado ################################################### + try: import re import socket @@ -279,10 +347,12 @@ except: logger.log("Could not import 'sliver_lxc' or 'libvirt' or 'sliver_libvirt'.") ################################################### + def get_sliver_ifconfig(slice_name, device="eth0"): - """ return the output of "ifconfig" run from inside the sliver. + """ + return the output of "ifconfig" run from inside the sliver. - side effects: adds "/usr/sbin" to sys.path + side effects: adds "/usr/sbin" to sys.path """ # See if setns is installed. If it's not then we're probably not running @@ -299,18 +369,20 @@ def get_sliver_ifconfig(slice_name, device="eth0"): if (not cgroup_fn) or (not pid): return None - path = '/proc/%s/ns/net'%pid + path = '/proc/{}/ns/net'.format(pid) result = None try: setns.chcontext(path) args = ["/sbin/ifconfig", device] - sub = subprocess.Popen(args, stderr = subprocess.PIPE, stdout = subprocess.PIPE) + 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()) + logger.log("get_slice_ifconfig: error in ifconfig: {}".format( + sub.stderr.read())) result = sub.stdout.read() finally: @@ -318,6 +390,7 @@ def get_sliver_ifconfig(slice_name, device="eth0"): return result + def get_sliver_ip(slice_name): ifconfig = get_sliver_ifconfig(slice_name) if not ifconfig: @@ -327,7 +400,7 @@ def get_sliver_ip(slice_name): 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:"): + if len(parts) >= 2 and parts[1].startswith("addr:"): return parts[1].split(":")[1] return None @@ -338,26 +411,31 @@ def get_sliver_ip(slice_name): # Get the slice ipv6 address # Only for LXC! ################################################### + + def get_sliver_ipv6(slice_name): ifconfig = get_sliver_ifconfig(slice_name) if not ifconfig: - return None,None + return None, None # example: 'inet6 2001:67c:16dc:1302:5054:ff:fea7:7882 prefixlen 64 scopeid 0x0' - prog = re.compile(r'inet6\s+(.*)\s+prefixlen\s+(\d+)\s+scopeid\s+(.+)') + prog = re.compile( + r'inet6\s+(.*)\s+prefixlen\s+(\d+)\s+scopeid\s+(.+)') for line in ifconfig.split("\n"): search = prog.search(line) if search: ipv6addr = search.group(1) prefixlen = search.group(2) - return (ipv6addr,prefixlen) - return None,None + return (ipv6addr, prefixlen) + return None, None ################################################### # Author: Guilherme Sperb Machado ################################################### # Check if the address is a AF_INET6 family address ################################################### + + def is_valid_ipv6(ipv6addr): try: socket.inet_pton(socket.AF_INET6, ipv6addr) @@ -365,33 +443,39 @@ def is_valid_ipv6(ipv6addr): return False return True -### this returns the kind of virtualization on the node + +# 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 (): +virt_stamp = "/etc/planetlab/virt" + + +def get_node_virt(): try: - return file(virt_stamp).read().strip() + with open(virt_stamp) as f: + return f.read().strip() except: pass logger.log("Computing virt..") try: - if subprocess.call ([ 'vserver', '--help' ]) ==0: virt='vs' - else: virt='lxc' + virt = 'vs' if subprocess.call(['vserver', '--help']) == 0 else 'lxc' except: - virt='lxc' - with file(virt_stamp,"w") as f: + virt = 'lxc' + with open(virt_stamp, "w") as f: f.write(virt) return virt -### this return True or False to indicate that systemctl is present on that box + +# 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 (): +_has_systemctl = None + + +def has_systemctl(): global _has_systemctl if _has_systemctl is None: - _has_systemctl = (subprocess.call([ 'systemctl', '--help' ]) == 0) + _has_systemctl = (subprocess.call(['systemctl', '--help']) == 0) return _has_systemctl ################################################### @@ -400,6 +484,8 @@ def has_systemctl (): # This method was developed to support the ipv6 plugin # Only for LXC! ################################################### + + def reboot_slivers(): type = 'sliver.LXC' # connecting to the libvirtd @@ -407,18 +493,41 @@ def reboot_slivers(): domains = connLibvirt.listAllDomains() for domain in domains: try: - domain.destroy() - logger.log("tools: DESTROYED %s" % (domain.name()) ) - domain.create() - logger.log("tools: CREATED %s" % (domain.name()) ) - except: - logger.log("tools: FAILED to reboot %s" % (domain.name()) ) + # set the flag VIR_DOMAIN_REBOOT_INITCTL, which uses "initctl" + result = domain.reboot(0x04) + if result == 0: + logger.log("tools: REBOOT {}".format(domain.name())) + else: + raise Exception() + except Exception as e: + logger.log("tools: FAILED to reboot {} ({})".format( + domain.name(), e)) + logger.log( + "tools: Trying to DESTROY/CREATE {} instead...".format(domain.name())) + try: + result = domain.destroy() + if result == 0: + logger.log("tools: DESTROYED {}".format(domain.name())) + else: + logger.log( + "tools: FAILED in the DESTROY call of {}".format(domain.name())) + result = domain.create() + if result == 0: + logger.log("tools: CREATED {}".format(domain.name())) + else: + logger.log( + "tools: FAILED in the CREATE call of {}".format(domain.name())) + except Exception as e: + logger.log( + "tools: FAILED to DESTROY/CREATE {} ({})".format(domain.name(), e)) ################################################### # Author: Guilherme Sperb Machado ################################################### # Get the /etc/hosts file path ################################################### + + def get_hosts_file_path(slicename): containerDir = os.path.join(sliver_lxc.Sliver_LXC.CON_BASE_DIR, slicename) return os.path.join(containerDir, 'etc', 'hosts') @@ -431,14 +540,16 @@ def get_hosts_file_path(slicename): # If the parameter 'ipv6addr' is None, then search # for any ipv6 address ################################################### + + def search_ipv6addr_hosts(slicename, ipv6addr): hostsFilePath = get_hosts_file_path(slicename) - found=False + found = False try: - for line in fileinput.input(r'%s' % (hostsFilePath)): + for line in fileinput.input(r'{}'.format(hostsFilePath)): if ipv6addr is not None: - if re.search(r'%s' % (ipv6addr), line): - found=True + if re.search(r'{}'.format(ipv6addr), line): + found = True else: search = re.search(r'^(.*)\s+.*$', line) if search: @@ -446,12 +557,12 @@ def search_ipv6addr_hosts(slicename, ipv6addr): ipv6candidatestrip = ipv6candidate.strip() valid = is_valid_ipv6(ipv6candidatestrip) if valid: - found=True - fileinput.close() - return found + found = True + fileinput.close() + return found except: - logger.log("tools: FAILED to search %s in /etc/hosts file of slice=%s" % \ - (ipv6addr, slicename) ) + logger.log("tools: FAILED to search {} in /etc/hosts file of slice={}" + .format(ipv6addr, slicename)) ################################################### # Author: Guilherme Sperb Machado @@ -459,32 +570,37 @@ def search_ipv6addr_hosts(slicename, ipv6addr): # Removes all ipv6 addresses from the /etc/hosts # file of a given slice ################################################### + + def remove_all_ipv6addr_hosts(slicename, node): hostsFilePath = get_hosts_file_path(slicename) try: - for line in fileinput.input(r'%s' % (hostsFilePath), inplace=True): - search = re.search(r'^(.*)\s+(%s|%s)$' % (node,'localhost'), line) + for line in fileinput.input(r'{}'.format(hostsFilePath), inplace=True): + search = re.search( + r'^(.*)\s+({}|{})$'.format(node, 'localhost'), line) if search: ipv6candidate = search.group(1) ipv6candidatestrip = ipv6candidate.strip() valid = is_valid_ipv6(ipv6candidatestrip) if not valid: - print line, + print(line, end=' ') fileinput.close() - logger.log("tools: REMOVED IPv6 address from /etc/hosts file of slice=%s" % \ - (slicename) ) + logger.log("tools: REMOVED IPv6 address from /etc/hosts file of slice={}" + .format(slicename)) except: - logger.log("tools: FAILED to remove the IPv6 address from /etc/hosts file of slice=%s" % \ - (slicename) ) + logger.log("tools: FAILED to remove the IPv6 address from /etc/hosts file of slice={}" + .format(slicename)) ################################################### # Author: Guilherme Sperb Machado ################################################### # Adds an ipv6 address to the /etc/hosts file within a slice ################################################### + + def add_ipv6addr_hosts_line(slicename, node, ipv6addr): hostsFilePath = get_hosts_file_path(slicename) - logger.log("tools: %s" % (hostsFilePath) ) + logger.log("tools: {}".format(hostsFilePath)) # debugging purposes: #string = "127.0.0.1\tlocalhost\n192.168.100.179\tmyplc-node1-vm.mgmt.local\n" #string = "127.0.0.1\tlocalhost\n" @@ -492,12 +608,11 @@ def add_ipv6addr_hosts_line(slicename, node, ipv6addr): with open(hostsFilePath, "a") as file: file.write(ipv6addr + " " + node + "\n") file.close() - logger.log("tools: ADDED IPv6 address to /etc/hosts file of slice=%s" % \ - (slicename) ) + logger.log("tools: ADDED IPv6 address to /etc/hosts file of slice={}" + .format(slicename)) except: - logger.log("tools: FAILED to add the IPv6 address to /etc/hosts file of slice=%s" % \ - (slicename) ) - + logger.log("tools: FAILED to add the IPv6 address to /etc/hosts file of slice={}" + .format(slicename)) # how to run a command in a slice @@ -509,22 +624,24 @@ def add_ipv6addr_hosts_line(slicename, node, ipv6addr): # 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': +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) ] + 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) + + +def init_signals(): + def handler(signum, frame): + logger.log("Received signal {} - exiting".format(signum)) os._exit(1) - signal.signal(signal.SIGHUP,handler) - signal.signal(signal.SIGQUIT,handler) - signal.signal(signal.SIGINT,handler) - signal.signal(signal.SIGTERM,handler) + signal.signal(signal.SIGHUP, handler) + signal.signal(signal.SIGQUIT, handler) + signal.signal(signal.SIGINT, handler) + signal.signal(signal.SIGTERM, handler)