X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=accounts.py;h=2f3bb329f6c9095246e0a02a8b0e2526c3ef854c;hb=refs%2Fheads%2Fdevel;hp=d540368deb21b2643fe8866e0986c40f16843775;hpb=737532f28237ed08d3895b5912866476f909e04f;p=nodemanager.git diff --git a/accounts.py b/accounts.py index d540368..2f3bb32 100644 --- a/accounts.py +++ b/accounts.py @@ -1,29 +1,37 @@ +### + """Functionality common to all account classes. -Each subclass of Account must provide 6 methods: create() and -destroy(), which are static; configure(), is_running(), start(), and stop(), -which are not. configure() takes a record as its only argument, and does -things like set up ssh keys. In addition, an Account subclass must +Each subclass of Account must provide five methods: + (*) create() and destroy(), which are static; + (*) configure(), start(), and stop(), which are not. + +configure(), which takes a record as its only argument, does +things like set up ssh keys. In addition, an Account subclass must provide static member variables SHELL, which contains the unique shell that it uses; and TYPE, a string that is used by the account creation code. For no particular reason, TYPE is divided hierarchically by periods; at the moment the only convention is that all sliver accounts have type that begins with sliver. +There are any number of race conditions that may result from the fact +that account names are not unique over time. Moreover, it's a bad +idea to perform lengthy operations while holding the database lock. +In order to deal with both of these problems, we use a worker thread +for each account name that ever exists. On 32-bit systems with large +numbers of accounts, this may cause the NM process to run out of +*virtual* memory! This problem may be remedied by decreasing the +maximum stack size. """ -import Queue import os -import pwd -from grp import getgrnam +import pwd, grp import threading import logger import tools -# When this variable is true, start after any ensure_created -Startingup = False # shell path -> account class association shell_acct_class = {} # account type -> account class association @@ -34,7 +42,9 @@ create_sem = threading.Semaphore(1) destroy_sem = threading.Semaphore(1) def register_class(acct_class): - """Call once for each account class. This method adds the class to the dictionaries used to look up account classes by shell and type.""" + """Call once for each account class. This method adds the class +to the dictionaries used to look up account classes by shell and +type.""" shell_acct_class[acct_class.SHELL] = acct_class type_acct_class[acct_class.TYPE] = acct_class @@ -61,74 +71,112 @@ def get(name): class Account: def __init__(self, rec): - logger.verbose('Initing account %s'%rec['name']) + logger.verbose('accounts: Initing account %s'%rec['name']) self.name = rec['name'] self.keys = '' - self.initscriptchanged = False self.configure(rec) @staticmethod def create(name, vref = None): abstract + @staticmethod def destroy(name): abstract def configure(self, rec): """Write to my authorized_keys file.""" - logger.verbose('%s: in accounts:configure'%self.name) + logger.verbose('accounts: configuring %s'%self.name) new_keys = rec['keys'] if new_keys != self.keys: - self.keys = new_keys - dot_ssh = '/home/%s/.ssh' % self.name - if not os.access(dot_ssh, os.F_OK): os.mkdir(dot_ssh) + # get the unix account info + gid = grp.getgrnam("slices")[2] + pw_info = pwd.getpwnam(self.name) + uid = pw_info[2] + pw_dir = pw_info[5] + + # write out authorized_keys file and conditionally create + # the .ssh subdir if need be. + dot_ssh = os.path.join(pw_dir,'.ssh') + if not os.path.isdir(dot_ssh): + if not os.path.isdir(pw_dir): + logger.verbose('accounts: WARNING: homedir %s does not exist for %s!'%(pw_dir,self.name)) + os.mkdir(pw_dir) + os.chown(pw_dir, uid, gid) + os.mkdir(dot_ssh) + + auth_keys = os.path.join(dot_ssh,'authorized_keys') + tools.write_file(auth_keys, lambda f: f.write(new_keys)) + + # set access permissions and ownership properly os.chmod(dot_ssh, 0700) - tools.write_file(dot_ssh + '/authorized_keys', lambda f: f.write(new_keys)) - logger.log('%s: installing ssh keys' % self.name) - user = pwd.getpwnam(self.name)[2] - group = getgrnam("slices")[2] - os.chown(dot_ssh, user, group) - os.chown(dot_ssh + '/authorized_keys', user, group) + os.chown(dot_ssh, uid, gid) + os.chmod(auth_keys, 0600) + os.chown(auth_keys, uid, gid) + + # set self.keys to new_keys only when all of the above ops succeed + self.keys = new_keys + + logger.log('accounts: %s: installed ssh keys' % self.name) def start(self, delay=0): pass def stop(self): pass def is_running(self): pass class Worker: + def __init__(self, name): self.name = name # username self._acct = None # the account object currently associated with this worker - def ensure_created(self, rec, startingup = Startingup): - """Check account type is still valid. If not, recreate sliver. If still valid, - check if running and configure/start if not.""" + def ensure_created(self, rec): + """Check account type is still valid. If not, recreate sliver. +If still valid, check if running and configure/start if not.""" + logger.log_data_in_file(rec,"/var/lib/nodemanager/%s.rec.txt"%rec['name'], + 'raw rec captured in ensure_created',logger.LOG_VERBOSE) curr_class = self._get_class() next_class = type_acct_class[rec['type']] if next_class != curr_class: self._destroy(curr_class) create_sem.acquire() - try: next_class.create(self.name, rec['vref']) + try: next_class.create(self.name, rec) finally: create_sem.release() if not isinstance(self._acct, next_class): self._acct = next_class(rec) - if startingup or \ - not self.is_running() or \ - next_class != curr_class or \ - self._acct.initscriptchanged: - self.start(rec) - else: self._acct.configure(rec) + logger.verbose("accounts.ensure_created: %s, running=%r"%(self.name,self.is_running())) + + # reservation_alive is set on reervable nodes, and its value is a boolean + if 'reservation_alive' in rec: + # reservable nodes + if rec['reservation_alive']: + # this sliver has the lease, it is safe to start it + if not self.is_running(): self.start(rec) + else: self.configure(rec) + else: + # not having the lease, do not start it + self.configure(rec) + # usual nodes - preserve old code + # xxx it's not clear what to do when a sliver changes type/class + # in a reservable node + else: + if not self.is_running() or next_class != curr_class: + self.start(rec) + else: self.configure(rec) def ensure_destroyed(self): self._destroy(self._get_class()) - def start(self, rec, d = 0): + def start(self, rec, d = 0): self._acct.configure(rec) self._acct.start(delay=d) + def configure(self, rec): + self._acct.configure(rec) + def stop(self): self._acct.stop() - def is_running(self): + def is_running(self): if (self._acct != None) and self._acct.is_running(): status = True else: status = False - logger.verbose("Worker(%s): is not running" % self.name) + logger.verbose("accounts: Worker(%s): is not running" % self.name) return status def _destroy(self, curr_class):