X-Git-Url: http://git.onelab.eu/?p=nodemanager.git;a=blobdiff_plain;f=database.py;h=a7099bc7cd0cb1717837a63cb5efdd1830229008;hp=1985b58da369c9b59c9238c8ad5546acc6af260e;hb=HEAD;hpb=33bb6d66fd29039576844c49ceceefe905d68fd3 diff --git a/database.py b/database.py index 1985b58..a7099bc 100644 --- a/database.py +++ b/database.py @@ -1,6 +1,4 @@ -# $Id$ -# $URL$ - +# """The database houses information on slivers. This information reaches the sliver manager in two different ways: one, through the GetSlivers() call made periodically; two, by users delivering tickets. @@ -15,26 +13,33 @@ In order to maintain service when the node reboots during a network partition, the database is constantly being dumped to disk. """ -import cPickle +import sys + +import pickle import threading import time -import accounts +import account import logger import tools import bwmon +# hopefully temporary +# is there a good reason to have this done here and not in a plugin ? +try: from coresched_lxc import CoreSched +except: from coresched_vs import CoreSched + # We enforce minimum allocations to keep the clueless from hosing their slivers. # Disallow disk loans because there's currently no way to punish slivers over quota. -MINIMUM_ALLOCATION = {'cpu_pct': 0, - 'cpu_share': 1, - 'net_min_rate': 0, - 'net_max_rate': 8, - 'net_i2_min_rate': 0, - 'net_i2_max_rate': 8, +MINIMUM_ALLOCATION = {'cpu_pct': 0, + 'cpu_share': 1, + 'net_min_rate': 0, + 'net_max_rate': 8, + 'net_i2_min_rate': 0, + 'net_i2_max_rate': 8, 'net_share': 1, } -LOANABLE_RESOURCES = MINIMUM_ALLOCATION.keys() +LOANABLE_RESOURCES = list(MINIMUM_ALLOCATION.keys()) DB_FILE = '/var/lib/nodemanager/database.pickle' @@ -64,19 +69,19 @@ class Database(dict): self._min_timestamp = 0 def _compute_effective_rspecs(self): - """Calculate the effects of loans and store the result in field _rspec. + """Calculate the effects of loans and store the result in field _rspec. At the moment, we allow slivers to loan only those resources that they have received directly from PLC. -In order to do the accounting, we store three different rspecs: - * field 'rspec', which is the resources given by PLC; - * field '_rspec', which is the actual amount of resources the sliver has after all loans; - * and variable resid_rspec, which is the amount of resources the sliver +In order to do the accounting, we store three different rspecs: + * field 'rspec', which is the resources given by PLC; + * field '_rspec', which is the actual amount of resources the sliver has after all loans; + * and variable resid_rspec, which is the amount of resources the sliver has after giving out loans but not receiving any.""" slivers = {} - for name, rec in self.iteritems(): + for name, rec in self.items(): if 'rspec' in rec: rec['_rspec'] = rec['rspec'].copy() slivers[name] = rec - for rec in slivers.itervalues(): + for rec in slivers.values(): eff_rspec = rec['_rspec'] resid_rspec = rec['rspec'].copy() for target, resource_name, amount in rec.get('_loans', []): @@ -93,18 +98,20 @@ keys.""" if rec['timestamp'] < self._min_timestamp: return name = rec['name'] old_rec = self.get(name) - if old_rec == None: self[name] = rec + if old_rec == None: + self[name] = rec elif rec['timestamp'] > old_rec['timestamp']: - for key in old_rec.keys(): - if not key.startswith('_'): del old_rec[key] + for key in list(old_rec.keys()): + if not key.startswith('_'): + del old_rec[key] old_rec.update(rec) def set_min_timestamp(self, ts): - """The ._min_timestamp member is the timestamp on the last comprehensive update. -We use it to determine if a record is stale. + """The ._min_timestamp member is the timestamp on the last comprehensive update. +We use it to determine if a record is stale. This method should be called whenever new GetSlivers() data comes in.""" self._min_timestamp = ts - for name, rec in self.items(): + for name, rec in list(self.items()): if rec['timestamp'] < ts: del self[name] def sync(self): @@ -115,40 +122,48 @@ It may be necessary in the future to do something smarter.""" # delete expired records now = time.time() - for name, rec in self.items(): + for name, rec in list(self.items()): if rec.get('expires', now) < now: del self[name] self._compute_effective_rspecs() + try: + coresched = CoreSched() + coresched.adjustCores(self) + except: + logger.log_exc("database: exception while doing core sched") + # create and destroy accounts as needed logger.verbose("database: sync : fetching accounts") - existing_acct_names = accounts.all() + existing_acct_names = account.all() for name in existing_acct_names: - if name not in self: + if name not in self: logger.verbose("database: sync : ensure_destroy'ing %s"%name) - accounts.get(name).ensure_destroyed() - for name, rec in self.iteritems(): - # protect this; if anything fails for a given sliver + account.get(name).ensure_destroyed() + for name, rec in self.items(): + # protect this; if anything fails for a given sliver # we still need the other ones to be handled try: - sliver = accounts.get(name) - logger.verbose("database: sync : looping on %s (shell account class from pwd %s)" %(name,sliver._get_class())) + sliver = account.get(name) + logger.verbose("database: sync : looping on %s (shell account class from pwd %s)" %(name, sliver._get_class())) # Make sure we refresh accounts that are running - if rec['instantiation'] == 'plc-instantiated': + if rec['instantiation'] == 'plc-instantiated': logger.verbose ("database: sync : ensure_create'ing 'instantiation' sliver %s"%name) sliver.ensure_created(rec) - elif rec['instantiation'] == 'nm-controller': + elif rec['instantiation'] == 'nm-controller': logger.verbose ("database: sync : ensure_create'ing 'nm-controller' sliver %s"%name) sliver.ensure_created(rec) # Back door to ensure PLC overrides Ticket in delegation. elif rec['instantiation'] == 'delegated' and sliver._get_class() != None: # if the ticket has been delivered and the nm-controller started the slice # update rspecs and keep them up to date. - if sliver.is_running(): + if sliver.is_running(): logger.verbose ("database: sync : ensure_create'ing 'delegated' sliver %s"%name) sliver.ensure_created(rec) + except SystemExit as e: + sys.exit(e) except: - logger.log_exc("database: sync failed to handle sliver",name=name) + logger.log_exc("database: sync failed to handle sliver", name=name) # Wake up bwmom to update limits. bwmon.lock.set() @@ -166,18 +181,19 @@ It proceeds to handle dump requests forever.""" while True: db_lock.acquire() while not dump_requested: db_cond.wait() - db_pickle = cPickle.dumps(db, cPickle.HIGHEST_PROTOCOL) + db_pickle = pickle.dumps(db, pickle.HIGHEST_PROTOCOL) dump_requested = False db_lock.release() - try: - tools.write_file(DB_FILE, lambda f: f.write(db_pickle)) + try: + tools.write_file( + DB_FILE, lambda f: f.write(db_pickle), binary=True) logger.log_database(db) - except: + except: logger.log_exc("database.start: failed to pickle/dump") global db try: f = open(DB_FILE) - try: db = cPickle.load(f) + try: db = pickle.load(f) finally: f.close() except IOError: logger.log ("database: Could not load %s -- starting from a fresh database"%DB_FILE)