X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=src%2Fnepi%2Fresources%2Flinux%2Fnode.py;h=d17e4dfb544577b67696240af263e912dd126b91;hb=4fea45903bd5ae406ea9965bdec3ada7119d0ea1;hp=9188022fa208e7a76041afaf253a187b1b19cd62;hpb=c957791ea6cdbd032f28cb62d4a39013dbfb7a4a;p=nepi.git diff --git a/src/nepi/resources/linux/node.py b/src/nepi/resources/linux/node.py index 9188022f..d17e4dfb 100644 --- a/src/nepi/resources/linux/node.py +++ b/src/nepi/resources/linux/node.py @@ -17,9 +17,9 @@ # # Author: Alina Quereilhac -from nepi.execution.attribute import Attribute, Flags -from nepi.execution.resource import ResourceManager, clsinit, ResourceState, \ - reschedule_delay +from nepi.execution.attribute import Attribute, Flags, Types +from nepi.execution.resource import ResourceManager, clsinit_copy, \ + ResourceState, reschedule_delay from nepi.resources.linux import rpmfuncs, debfuncs from nepi.util import sshfuncs, execfuncs from nepi.util.sshfuncs import ProcStatus @@ -57,7 +57,7 @@ class OSType: UBUNTU = "ubuntu" DEBIAN = "debian" -@clsinit +@clsinit_copy class LinuxNode(ResourceManager): """ .. class:: Class Args : @@ -142,6 +142,9 @@ class LinuxNode(ResourceManager): """ _rtype = "LinuxNode" + _help = "Controls Linux host machines ( either localhost or a host " \ + "that can be accessed using a SSH key)" + _backend_type = "linux" @classmethod def _register_attributes(cls): @@ -165,14 +168,20 @@ class LinuxNode(ResourceManager): clean_home = Attribute("cleanHome", "Remove all nepi files and directories " " from node home folder before starting experiment", + type = Types.Bool, + default = False, flags = Flags.ExecReadOnly) clean_experiment = Attribute("cleanExperiment", "Remove all files and directories " " from a previous same experiment, before the new experiment starts", + type = Types.Bool, + default = False, flags = Flags.ExecReadOnly) clean_processes = Attribute("cleanProcesses", "Kill all running processes before starting experiment", + type = Types.Bool, + default = False, flags = Flags.ExecReadOnly) tear_down = Attribute("tearDown", "Bash script to be executed before " + \ @@ -196,8 +205,14 @@ class LinuxNode(ResourceManager): # home directory at Linux host self._home_dir = "" - # lock to avoid concurrency issues on methods used by applications - self._lock = threading.Lock() + # lock to prevent concurrent applications on the same node, + # to execute commands at the same time. There are potential + # concurrency issues when using SSH to a same host from + # multiple threads. There are also possible operational + # issues, e.g. an application querying the existence + # of a file or folder prior to its creation, and another + # application creating the same file or folder in between. + self._node_lock = threading.Lock() def log_message(self, msg): return " guid %d - host %s - %s " % (self.guid, @@ -264,6 +279,8 @@ class LinuxNode(ResourceManager): self._os = OSType.FEDORA_12 elif out.find("Fedora release 14") == 0: self._os = OSType.FEDORA_14 + elif out.find("Fedora release") == 0: + self._os = OSType.FEDORA elif out.find("Debian") == 0: self._os = OSType.DEBIAN elif out.find("Ubuntu") ==0: @@ -281,25 +298,16 @@ class LinuxNode(ResourceManager): # To work arround this, repeat the operation N times or # until the result is not empty string out = "" - retrydelay = 1.0 - for i in xrange(10): - try: - (out, err), proc = self.execute("cat /etc/issue", - retry = 5, - with_lock = True, - blocking = True) - - if out.strip() != "": - return out - except: - trace = traceback.format_exc() - msg = "Error detecting OS: %s " % trace - self.error(msg, out, err) - return False - - time.sleep(min(30.0, retrydelay)) - retrydelay *= 1.5 - + try: + (out, err), proc = self.execute("cat /etc/issue", + with_lock = True, + blocking = True) + except: + trace = traceback.format_exc() + msg = "Error detecting OS: %s " % trace + self.error(msg, out, err) + + return out @property def use_deb(self): @@ -314,11 +322,9 @@ class LinuxNode(ResourceManager): def localhost(self): return self.get("hostname") in ['localhost', '127.0.0.7', '::1'] - def provision(self): + def do_provision(self): # check if host is alive if not self.is_alive(): - self.fail() - msg = "Deploy failed. Unresponsive node %s" % self.get("hostname") self.error(msg) raise RuntimeError, msg @@ -343,34 +349,31 @@ class LinuxNode(ResourceManager): # Create experiment node home directory self.mkdir(self.node_home) - super(LinuxNode, self).provision() + super(LinuxNode, self).do_provision() - def deploy(self): + def do_deploy(self): if self.state == ResourceState.NEW: - try: - self.discover() - self.provision() - except: - self._state = ResourceState.FAILED - raise + self.info("Deploying node") + self.do_discover() + self.do_provision() # Node needs to wait until all associated interfaces are # ready before it can finalize deployment from nepi.resources.linux.interface import LinuxInterface - ifaces = self.get_connected(LinuxInterface.rtype()) + ifaces = self.get_connected(LinuxInterface.get_rtype()) for iface in ifaces: if iface.state < ResourceState.READY: self.ec.schedule(reschedule_delay, self.deploy) return - super(LinuxNode, self).deploy() + super(LinuxNode, self).do_deploy() - def release(self): - # Node needs to wait until all associated RMs are released - # to be released + def do_release(self): rms = self.get_connected() for rm in rms: - if rm.state < ResourceState.STOPPED: + # Node needs to wait until all associated RMs are released + # before it can be released + if rm.state != ResourceState.RELEASED: self.ec.schedule(reschedule_delay, self.release) return @@ -380,32 +383,22 @@ class LinuxNode(ResourceManager): self.clean_processes() - super(LinuxNode, self).release() + super(LinuxNode, self).do_release() def valid_connection(self, guid): # TODO: Validate! return True - def clean_processes(self, killer = False): + def clean_processes(self): self.info("Cleaning up processes") - if killer: - # Hardcore kill - cmd = ("sudo -S killall python tcpdump || /bin/true ; " + - "sudo -S killall python tcpdump || /bin/true ; " + - "sudo -S kill $(ps -N -T -o pid --no-heading | grep -v $PPID | sort) || /bin/true ; " + - "sudo -S killall -u root || /bin/true ; " + - "sudo -S killall -u root || /bin/true ; ") - else: - # Be gentler... - cmd = ("sudo -S killall tcpdump || /bin/true ; " + - "sudo -S killall tcpdump || /bin/true ; " + - "sudo -S killall -u %s || /bin/true ; " % self.get("username") + - "sudo -S killall -u %s || /bin/true ; " % self.get("username")) + cmd = ("sudo -S killall tcpdump || /bin/true ; " + + "sudo -S kill $(ps aux | grep '[n]epi' | awk '{print $2}') || /bin/true ; " + + "sudo -S killall -u %s || /bin/true ; " % self.get("username")) out = err = "" - (out, err), proc = self.execute(cmd, retry = 1, with_lock = True) - + (out, err), proc = self.execute(cmd, retry = 1, with_lock = True) + def clean_home(self): """ Cleans all NEPI related folders in the Linux host """ @@ -456,7 +449,7 @@ class LinuxNode(ResourceManager): env = env) else: if with_lock: - with self._lock: + with self._node_lock: (out, err), proc = sshfuncs.rexec( command, host = self.get("hostname"), @@ -524,7 +517,7 @@ class LinuxNode(ResourceManager): sudo = sudo, user = user) else: - with self._lock: + with self._node_lock: (out, err), proc = sshfuncs.rspawn( command, pidfile = pidfile, @@ -549,7 +542,7 @@ class LinuxNode(ResourceManager): if self.localhost: pidtuple = execfuncs.lgetpid(os.path.join(home, pidfile)) else: - with self._lock: + with self._node_lock: pidtuple = sshfuncs.rgetpid( os.path.join(home, pidfile), host = self.get("hostname"), @@ -566,7 +559,7 @@ class LinuxNode(ResourceManager): if self.localhost: status = execfuncs.lstatus(pid, ppid) else: - with self._lock: + with self._node_lock: status = sshfuncs.rstatus( pid, ppid, host = self.get("hostname"), @@ -588,7 +581,7 @@ class LinuxNode(ResourceManager): if self.localhost: (out, err), proc = execfuncs.lkill(pid, ppid, sudo) else: - with self._lock: + with self._node_lock: (out, err), proc = sshfuncs.rkill( pid, ppid, host = self.get("hostname"), @@ -608,7 +601,7 @@ class LinuxNode(ResourceManager): recursive = True, strict_host_checking = False) else: - with self._lock: + with self._node_lock: (out, err), proc = sshfuncs.rcopy( src, dst, port = self.get("port"), @@ -619,7 +612,6 @@ class LinuxNode(ResourceManager): return (out, err), proc - def upload(self, src, dst, text = False, overwrite = True): """ Copy content to destination @@ -901,7 +893,7 @@ class LinuxNode(ResourceManager): pid = ppid = None delay = 1.0 - for i in xrange(4): + for i in xrange(2): pidtuple = self.getpid(home = home, pidfile = pidfile) if pidtuple: @@ -953,35 +945,25 @@ class LinuxNode(ResourceManager): return True out = err = "" + msg = "Unresponsive host. Wrong answer. " + # The underlying SSH layer will sometimes return an empty # output (even if the command was executed without errors). # To work arround this, repeat the operation N times or # until the result is not empty string - retrydelay = 1.0 - for i in xrange(10): - try: - (out, err), proc = self.execute("echo 'ALIVE'", - retry = 5, - blocking = True, - with_lock = True) - - if out.find("ALIVE") > -1: - return True - except: - trace = traceback.format_exc() - msg = "Unresponsive host. Error reaching host: %s " % trace - self.error(msg, out, err) - return False - - time.sleep(min(30.0, retrydelay)) - retrydelay *= 1.5 + try: + (out, err), proc = self.execute("echo 'ALIVE'", + blocking = True, + with_lock = True) + + if out.find("ALIVE") > -1: + return True + except: + trace = traceback.format_exc() + msg = "Unresponsive host. Error reaching host: %s " % trace - if out.find("ALIVE") > -1: - return True - else: - msg = "Unresponsive host. Wrong answer. " - self.error(msg, out, err) - return False + self.error(msg, out, err) + return False def find_home(self): """ Retrieves host home directory @@ -990,28 +972,19 @@ class LinuxNode(ResourceManager): # output (even if the command was executed without errors). # To work arround this, repeat the operation N times or # until the result is not empty string - retrydelay = 1.0 - for i in xrange(10): - try: - (out, err), proc = self.execute("echo ${HOME}", - retry = 5, - blocking = True, - with_lock = True) - - if out.strip() != "": - self._home_dir = out.strip() - break - except: - trace = traceback.format_exc() - msg = "Impossible to retrieve HOME directory" % trace - self.error(msg, out, err) - return False - - time.sleep(min(30.0, retrydelay)) - retrydelay *= 1.5 + msg = "Impossible to retrieve HOME directory" + try: + (out, err), proc = self.execute("echo ${HOME}", + blocking = True, + with_lock = True) + + if out.strip() != "": + self._home_dir = out.strip() + except: + trace = traceback.format_exc() + msg = "Impossible to retrieve HOME directory" % trace if not self._home_dir: - msg = "Impossible to retrieve HOME directory" self.error(msg, out, err) raise RuntimeError, msg