X-Git-Url: http://git.onelab.eu/?a=blobdiff_plain;f=PLC%2FMethods%2FRefreshPeer.py;h=7af8569f84fc1035585cd60029e200f708a59023;hb=19d4a01ccf66af9e00914351b3eacd5fc880f988;hp=8f4aaef7103bf3050ca15c8e716e784a222c0123;hpb=2711ea694a427f17256f28d36bec6b9afc6c0575;p=plcapi.git diff --git a/PLC/Methods/RefreshPeer.py b/PLC/Methods/RefreshPeer.py index 8f4aaef..7af8569 100644 --- a/PLC/Methods/RefreshPeer.py +++ b/PLC/Methods/RefreshPeer.py @@ -1,74 +1,879 @@ # # Thierry Parmentelat - INRIA -# - -import xmlrpclib +# +import os +import sys +import fcntl +import time +from PLC.Logger import logger from PLC.Faults import * from PLC.Method import Method from PLC.Parameter import Parameter, Mixed from PLC.Auth import Auth from PLC.Peers import Peer, Peers +from PLC.Sites import Site, Sites from PLC.Persons import Person, Persons -##from PLC.ForeignNodes import ForeignNode, ForeignNodes +from PLC.KeyTypes import KeyType, KeyTypes +from PLC.Keys import Key, Keys +from PLC.BootStates import BootState, BootStates +from PLC.Nodes import Node, Nodes +from PLC.SliceInstantiations import SliceInstantiations +from PLC.Slices import Slice, Slices +from PLC.Roles import Role, Roles + +# settings +# initial version was doing only one final commit +# * set commit_mode to False to get that behaviour +# * set comit_mode to True to get everything synced at once +# the issue with the 'one-commit-at-the-end' approach is +# that the db gets basically totally locked during too long +# causing various issues/crashes in the rest of the system +commit_mode = True + +# turn this to False only if both ends have the same db schema +# compatibility mode is a bit slower but probably safer on the long run +compatibility = True + +# debugging +# for verbose output +verbose = False +use_cache = None +# for debugging specific entries - display detailed info on selected objs +focus_type = None # set to e.g. 'Person' +# set to a list of ids (e.g. person_ids) - remote or local ids should work +focus_ids = [] +# DEBUGGING SETTINGS example +# set to a filename for using cached data when debugging +# WARNING: does *not* actually connect to the peer in this case +# or more precisely, connect only if the file is not found +# i.e. the first time +# use_cache = "/var/log/peers/getpeerdata.json" +# verbose = True +# focus_type = 'Person' +# focus_ids = [621, 1088] + + +########## helpers + +def message(to_print=None, verbose_only=False): + if verbose_only and not verbose: + return + logger.info(to_print) + + +def message_verbose(to_print=None, header='VERBOSE'): + message("{}> {}".format(header, to_print), verbose_only=True) + + +# to avoid several instances running at the same time +class FileLock: + """ + Lock/Unlock file + """ + + def __init__(self, file_path, expire=60 * 60 * 2): + self.expire = expire + self.fpath = file_path + self.fd = None + + def lock(self): + if os.path.exists(self.fpath): + if (time.time() - os.stat(self.fpath).st_ctime) > self.expire: + try: + os.unlink(self.fpath) + except Exception, e: + message('FileLock.lock({}) : {}'.format(self.fpath, e)) + return False + try: + self.fd = open(self.fpath, 'w') + fcntl.flock(self.fd, fcntl.LOCK_EX | fcntl.LOCK_NB) + except IOError, e: + message('FileLock.lock({}) : {}'.format(self.fpath, e)) + return False + return True + + def unlock(self): + try: + fcntl.flock(self.fd, fcntl.LOCK_UN | fcntl.LOCK_NB) + self.fd.close() + except IOError, e: + message('FileLock.unlock({}) : {}'.format(self.fpath, e)) class RefreshPeer(Method): """ - Query a peer PLC for its list of nodes, and refreshes - the local database accordingly - - Returns a dict containing - (*) 'plcname' : the peer name - (*) 'new_nodes' : the number of new nodes from that peer - may be negative - (*) 'new_slices': the number of new slices from that peer - may be negative + Fetches site, node, slice, person and key data from the specified peer + and caches it locally; also deletes stale entries. + Upon successful completion, returns a dict reporting various timers. + Faults otherwise. """ - + roles = ['admin'] - - accepts = [ Auth(), - Parameter (int, "Peer id") ] - - returns = Parameter(dict, "plcname, new_nodes, new_slices") - - def call (self, auth, peer_id): - - ### retrieve peer info - peers = Peers (self.api,[peer_id]) - try: - peer=peers[0] - except: - raise PLCInvalidArgument,'no such peer_id:%d'%peer_id - - ### retrieve account info - person_id = peer['person_id'] - persons = Persons (self.api,[person_id]) + + accepts = [ + Auth(), + Mixed(Peer.fields['peer_id'], + Peer.fields['peername']), + ] + + returns = Parameter(dict, "various timers") + + # the columns ignored in the comparison - it is safe to ignore: + # (*) the primary key - obviously this is not expected to match + # (*) peer_id and the transcode key, likewise + # (*) all identifiers that refer to other objects in the db + # like e.g. Person.site_ids since this is managed separately + # and does not make sense any way + # (*) we also ignore things like date_created and last_updated + # that refer to local db creation/update + ignore_site_fields = [ + 'site_id', 'peer_id', 'peer_site_id', + 'address_ids', 'node_ids', 'person_ids', 'pcu_ids', 'slice_ids', 'site_tag_ids', + 'date_created', 'last_updated', + ] + ignore_key_fields = [ + 'key_id', 'peer_id', 'peer_key_id', + 'person_id', + ] + ignore_person_fields = [ + 'person_id', 'peer_id', 'peer_person_id', + 'key_ids', 'slice_ids', 'person_tag_ids', 'role_ids', 'roles', 'site_ids', + 'date_created', 'last_updated', + ] + ignore_node_fields = [ + 'node_id', 'peer_id', 'peer_node_id', + 'node_tag_ids', 'interface_ids', 'slice_ids', 'nodegroup_ids', 'pcu_ids', 'ports', + 'date_created', 'last_updated', + # somehow those won't print in the ple db + 'last_download', 'last_contact', 'last_pcu_reboot', 'last_boot', + 'last_time_spent_offline', 'last_time_spent_online', 'last_pcu_confirmation', + ] + ignore_slice_fields = [ + 'slice_id', 'peer_id', 'peer_slice_id', + 'person_ids', 'slice_tag_ids', 'node_ids', + 'created', + ] + + def call(self, auth, peer_id_or_peername): + ret_val = None + peername = Peers(self.api, [peer_id_or_peername], [ + 'peername'])[0]['peername'] + file_lock = FileLock("/tmp/refresh-peer-{peername}.lock" + .format(peername=peername)) + if not file_lock.lock(): + raise Exception, "Another instance of RefreshPeer is running." try: - person = persons[0] - except: - raise PLCInvalidArgument,'no such person_id:%d'%person_id - - ### build up foreign auth - auth={ 'Username': person['email'], - 'AuthMethod' : 'password', - 'AuthString' : person['password'], - 'Role' : 'admin' } - - ## connect to the peer's API - url=peer['peer_url'] - print 'url=',url - apiserver = xmlrpclib.Server (url) - print 'auth=',auth - - peer_get_nodes = apiserver.GetNodes(auth) - nb_new_nodes = peer.refresh_nodes(peer_get_nodes) - - # rough and temporary - peer_foreign_nodes = apiserver.GetForeignNodes(auth) - peer_get_slices = apiserver.GetSlices(auth) - nb_new_slices = peer.refresh_slices(peer_get_slices,peer_foreign_nodes) - - return {'plcname':self.api.config.PLC_NAME, - 'new_nodes':nb_new_nodes, - 'new_slices':nb_new_slices} + ret_val = self.real_call(auth, peer_id_or_peername) + except Exception, e: + file_lock.unlock() + logger.exception("RefreshPeer caught exception - BEG") + message("RefreshPeer caught exception - END") + raise Exception, e + file_lock.unlock() + return ret_val + + def real_call(self, auth, peer_id_or_peername): + # Get peer + peers = Peers(self.api, [peer_id_or_peername]) + if not peers: + raise PLCInvalidArgument, "No such peer '{}'".format(unicode(peer_id_or_peername)) + peer = peers[0] + peer_id = peer['peer_id'] + peername = peer['peername'] + + # Connect to peer API + peer.connect() + + timers = {} + + # Get peer data + start = time.time() + message('RefreshPeer starting up (commit_mode={})'.format(commit_mode)) + if not use_cache: + message('Issuing GetPeerData') + peer_tables = peer.GetPeerData() + else: + import json + if os.path.isfile(use_cache): + message("use_cache: WARNING: using cached getpeerdata") + with open(use_cache) as storage: + peer_tables = json.load(storage) + else: + message("use_cache: issuing GetPeerData") + peer_tables = peer.GetPeerData() + message("use_cache: saving in cache {}".format(use_cache)) + with open(use_cache, 'w') as storage: + json.dump(peer_tables, storage) + + # additions in June 2017 + + # remove entries not marked as enabled + # actually the 'enabled' flag is present on 'Sites' and 'Persons' + # however we accept disabled slices as + # (*) they don't come and go too often + # (*) they may contain vlid nodes, that we would then lose + # if we were to discard those sites + # so bottom line, we filter out only disabled persons + for cls in ('Persons',) : + peer_tables[cls] = [ + obj for obj in peer_tables[cls] if obj['enabled'] + ] + + # somehow we can see GetPeerData from PLC that contains references + # to nodes that are not exposed themselves + # which suggests some inconsistency on their end + # anyway, it's safer to sanitize the dataset to avoid corruption + exposed_peer_node_ids = { n['node_id'] for n in peer_tables['Nodes']} + for slice in peer_tables['Slices']: + before = len(slice['node_ids']) + slice['node_ids'] = [x for x in slice['node_ids'] if x in exposed_peer_node_ids] + after = len(slice['node_ids']) + if after != before: + message("{peername} slice {slicename} got sanitized - {diff} node entries removed out of {before}" + .format(peername=peername, slicename=slice['name'], + diff=before-after, before=before)) + + # end of additions + + # for smooth federation with 4.2 - ignore fields that are useless + # anyway, and rewrite boot_state + boot_state_rewrite = {'dbg': 'safeboot', 'diag': 'safeboot', 'disable': 'disabled', + 'inst': 'reinstall', 'rins': 'reinstall', 'new': 'reinstall', 'rcnf': 'reinstall'} + for node in peer_tables['Nodes']: + for key in ['nodenetwork_ids', 'dummybox_id']: + if key in node: + del node[key] + if node['boot_state'] in boot_state_rewrite: + node['boot_state'] = boot_state_rewrite[node['boot_state']] + for slice in peer_tables['Slices']: + for key in ['slice_attribute_ids']: + if key in slice: + del slice[key] + timers['transport'] = time.time() - start - peer_tables['db_time'] + timers['peer_db'] = peer_tables['db_time'] + message_verbose('GetPeerData returned -> db={} transport={}' + .format(timers['peer_db'], timers['transport'])) + + def sync(objects, peer_objects, classobj, columns): + """ + Synchronizes two dictionaries of objects. objects should + be a dictionary of local objects keyed on their foreign + identifiers. peer_objects should be a dictionary of + foreign objects keyed on their local (i.e., foreign to us) + identifiers. Returns a final dictionary of local objects + keyed on their foreign identifiers. + """ + + classname = classobj(self.api).__class__.__name__ + primary_key = getattr(classobj, 'primary_key') + # display all peer objects of these types while looping + secondary_keys = {'Node': 'hostname', 'Slice': 'name', + 'Person': 'email', 'Site': 'login_base'} + secondary_key = None + if classname in secondary_keys: + secondary_key = secondary_keys[classname] + + message_verbose('Entering sync on {} ({})' + .format(classname, primary_key)) + + synced = {} + + # Delete stale objects + for peer_object_id, object in objects.iteritems(): + if peer_object_id not in peer_objects: + object.delete(commit=commit_mode) + message("{} {} {} deleted" + .format(peername, classname, object[primary_key])) + + total = len(peer_objects) + count = 1 + + # peer_object_id, peer_object and object are open variables in the loop below... + # (local) object might be None if creating a new one + def in_focus(): + if classname != focus_type: + return False + return (peer_object_id in focus_ids) \ + or (object and (primary_key in object) + and (object[primary_key] in focus_ids)) + + def message_focus(message): + if in_focus(): + # always show remote + message_verbose("peer_obj : {} [[{}]]".format(peer_object_id, peer_object), + header='FOCUS ' + message) + # show local object if a match was found + if object: + message_verbose("local_obj : <<{}>>".format(object), + header='FOCUS ' + message) + + # the function to compare a local object with its candidate peer obj + # xxx probably faster when compatibility is False... + def equal_fields(object, peer_object, columns): + # fast version: must use __eq__() instead of == since + # peer_object may be a raw dict instead of a Peer object. + if not compatibility: + result = object.__eq__(peer_object) + if not result: + message_verbose("fast mode: difference found between {} and {}" + .format(object, peer_object)) + return result + else: + for column in columns: + if object[column] != peer_object[column]: + message_verbose("difference found in column {}".format(column)) + message_verbose("our object {}".format(object[column])) + message_verbose("remote object {}".format(peer_object[column])) + return False + return True + + # Add/update new/existing objects + for peer_object_id, peer_object in peer_objects.iteritems(): + peer_object_name = "" + if secondary_key: + peer_object_name = "({})".format(peer_object[secondary_key]) + message_verbose('{} peer_object_id={} {} ({}/{})' + .format(classname, peer_object_id, peer_object_name, count, total)) + count += 1 + if peer_object_id in synced: + message("Warning: {peername} Skipping already added {classname}: {obj}" + .format(peername=peername, + classname=classname, obj=peer_object)) + continue + + if peer_object_id in objects: + # Update existing object + object = objects[peer_object_id] + + # Replace foreign identifier with existing local + # identifier temporarily for the purposes of + # comparison. + peer_object[primary_key] = object[primary_key] + + if not equal_fields(object, peer_object, columns): + # Only update intrinsic fields + object.update(object.db_fields(peer_object)) + message_focus("DIFFERENCES : updated / syncing") + sync = True + action = "changed" + else: + message_focus("UNCHANGED - left intact / not syncing") + sync = False + action = None + + # Restore foreign identifier + peer_object[primary_key] = peer_object_id + else: + object = None + # Add new object + object = classobj(self.api, peer_object) + # Replace foreign identifier with new local identifier + del object[primary_key] + message_focus("NEW -- created with clean id - syncing") + sync = True + action = "added" + + if sync: + message_verbose("syncing {classname} {id} - commit_mode={mode}" + .format(classname=classname, + id=peer_object_id, mode=commit_mode)) + try: + object.sync(commit=commit_mode) + except PLCInvalidArgument, err: + # XXX Log an event instead of printing to logfile + # skip if validation fails + message("Warning: {peername} Skipping invalid {classname} ({err})\n{object}" + .format(peername=peername, classname=classname, + object=peer_object, err=err)) + continue + + synced[peer_object_id] = object + + if action: + message("{peername}: ({count}/{total}) {classname} {primary} {name} {action}" + .format(peername=peername, + count=count, total=total, + classname=classname, primary=object[primary_key], + name=peer_object_name, action=action)) + + message_verbose("Exiting sync on {}".format(classname)) + + return synced + + # over time, we've had issues with a given column being + # added on one side and not on the other + # this helper function computes the intersection of two list of + # fields/columns + def intersect(l1, l2): + if compatibility: + return list(set(l1).intersection(set(l2))) + else: + return l1 + + # some fields definitely need to be ignored + def ignore(l1, l2): + return list(set(l1).difference(set(l2))) + + # + # Synchronize foreign sites + # + + start = time.time() + + message('(1) Dealing with Sites') + + # Compare only the columns returned by the GetPeerData() call + if peer_tables['Sites']: + columns = peer_tables['Sites'][0].keys() + columns = intersect(columns, Site.fields) + else: + columns = None + + # Keyed on foreign site_id + old_peer_sites = Sites( + self.api, {'peer_id': peer_id}, columns).dict('peer_site_id') + sites_at_peer = dict([(site['site_id'], site) + for site in peer_tables['Sites']]) + + # Synchronize new set (still keyed on foreign site_id) + peer_sites = sync(old_peer_sites, sites_at_peer, Site, + ignore(columns, RefreshPeer.ignore_site_fields)) + + for peer_site_id, site in peer_sites.iteritems(): + # Bind any newly cached sites to peer + if peer_site_id not in old_peer_sites: + peer.add_site(site, peer_site_id, commit=commit_mode) + site['peer_id'] = peer_id + site['peer_site_id'] = peer_site_id + + timers['site'] = time.time() - start + + # + # XXX Synchronize foreign key types + # + + message('(2) Dealing with Keys') + + key_types = KeyTypes(self.api).dict() + + # + # Synchronize foreign keys + # + + start = time.time() + + # Compare only the columns returned by the GetPeerData() call + if peer_tables['Keys']: + columns = peer_tables['Keys'][0].keys() + columns = intersect(columns, Key.fields) + else: + columns = None + + # Keyed on foreign key_id + old_peer_keys = Keys( + self.api, {'peer_id': peer_id}, columns).dict('peer_key_id') + keys_at_peer = dict([(key['key_id'], key) + for key in peer_tables['Keys']]) + + # Fix up key_type references + for peer_key_id, key in keys_at_peer.items(): + if key['key_type'] not in key_types: + # XXX Log an event instead of printing to logfile + message("Warning: Skipping invalid {peername} key {key}" + .format(peername=peername, key=key)) + del keys_at_peer[peer_key_id] + continue + + # Synchronize new set (still keyed on foreign key_id) + peer_keys = sync(old_peer_keys, keys_at_peer, Key, + ignore(columns, RefreshPeer.ignore_key_fields)) + for peer_key_id, key in peer_keys.iteritems(): + # Bind any newly cached keys to peer + if peer_key_id not in old_peer_keys: + peer.add_key(key, peer_key_id, commit=commit_mode) + key['peer_id'] = peer_id + key['peer_key_id'] = peer_key_id + + timers['keys'] = time.time() - start + + # + # Synchronize foreign users + # + + start = time.time() + + message('(3) Dealing with Persons') + + # Compare only the columns returned by the GetPeerData() call + if peer_tables['Persons']: + columns = peer_tables['Persons'][0].keys() + columns = intersect(columns, Person.fields) + else: + columns = None + + # Keyed on foreign person_id + old_peer_persons = Persons( + self.api, {'peer_id': peer_id}, columns).dict('peer_person_id') + + # artificially attach the persons returned by GetPeerData to the new peer + # this is because validate_email needs peer_id to be correct when + # checking for duplicates + for person in peer_tables['Persons']: + person['peer_id'] = peer_id + persons_at_peer = dict([(peer_person['person_id'], peer_person) + for peer_person in peer_tables['Persons']]) + + # XXX Do we care about membership in foreign site(s)? + + # Synchronize new set (still keyed on foreign person_id) + peer_persons = sync(old_peer_persons, persons_at_peer, Person, + ignore(columns, RefreshPeer.ignore_person_fields)) + + # transcoder : retrieve a local key_id from a peer_key_id + key_transcoder = dict([(key['key_id'], peer_key_id) + for peer_key_id, key in peer_keys.iteritems()]) + + for peer_person_id, person in peer_persons.iteritems(): + # Bind any newly cached users to peer + if peer_person_id not in old_peer_persons: + peer.add_person(person, peer_person_id, commit=commit_mode) + person['peer_id'] = peer_id + person['peer_person_id'] = peer_person_id + person['key_ids'] = [] + + # User as viewed by peer + peer_person = persons_at_peer[peer_person_id] + + # Foreign keys currently belonging to the user + old_person_key_ids = [key_transcoder[key_id] for key_id in person['key_ids'] + if key_transcoder[key_id] in peer_keys] + + # Foreign keys that should belong to the user + # this is basically peer_person['key_ids'], we just check it makes sense + # (e.g. we might have failed importing it) + person_key_ids = [key_id for key_id in peer_person[ + 'key_ids'] if key_id in peer_keys] + + # Remove stale keys from user + for key_id in (set(old_person_key_ids) - set(person_key_ids)): + person.remove_key(peer_keys[key_id], commit=commit_mode) + message("{peername} Key {key_id} removed from person {email}" + .format(peername=peername, + key_id=key_id, email=person['email'])) + + # Add new keys to user + for key_id in (set(person_key_ids) - set(old_person_key_ids)): + #message("before add_key, passing person={}".format(person)) + #message("before add_key, passing key={}".format(peer_keys[key_id])) + person.add_key(peer_keys[key_id], commit=commit_mode) + message("{} Key {} added into person {}" + .format(peername, key_id, person['email'])) + + timers['persons'] = time.time() - start + + # + # XXX Synchronize foreign boot states + # + + boot_states = BootStates(self.api).dict() + + # + # Synchronize foreign nodes + # + + start = time.time() + + # NOTE: we do import disabled sites + message('(4) Dealing with Nodes (1)') + + # Compare only the columns returned by the GetPeerData() call + if peer_tables['Nodes']: + columns = peer_tables['Nodes'][0].keys() + columns = intersect(columns, Node.fields) + else: + columns = Node.fields + + # Keyed on foreign node_id + old_peer_nodes = Nodes( + self.api, {'peer_id': peer_id}, columns).dict('peer_node_id') + nodes_at_peer = dict([(node['node_id'], node) + for node in peer_tables['Nodes']]) + + # Fix up site_id and boot_states references + for peer_node_id, node in nodes_at_peer.items(): + errors = [] + if node['site_id'] not in peer_sites: + errors.append("invalid (or disabled) site {}".format(node['site_id'])) + if node['boot_state'] not in boot_states: + errors.append("invalid boot state {}".format(node['boot_state'])) + if errors: + # XXX Log an event instead of printing to logfile + message("Warning: Skipping invalid {peername} node {hostname} - {errors}" + .format(peername=peername, + hostname=node['hostname'], errors=", ".join(errors))) + del nodes_at_peer[peer_node_id] + continue + else: + node['site_id'] = peer_sites[node['site_id']]['site_id'] + + # Synchronize new set + peer_nodes = sync(old_peer_nodes, nodes_at_peer, Node, + ignore(columns, RefreshPeer.ignore_node_fields)) + + for peer_node_id, node in peer_nodes.iteritems(): + # Bind any newly cached foreign nodes to peer + if peer_node_id not in old_peer_nodes: + peer.add_node(node, peer_node_id, commit=commit_mode) + node['peer_id'] = peer_id + node['peer_node_id'] = peer_node_id + + timers['nodes'] = time.time() - start + + # + # Synchronize local nodes + # + + start = time.time() + message('(5) Dealing with Nodes (2)') + + # Keyed on local node_id + local_nodes = Nodes(self.api).dict() + + for node in peer_tables['PeerNodes']: + # Foreign identifier for our node as maintained by peer + peer_node_id = node['node_id'] + # Local identifier for our node as cached by peer + node_id = node['peer_node_id'] + if node_id in local_nodes: + # Still a valid local node, add it to the synchronized + # set of local node objects keyed on foreign node_id. + peer_nodes[peer_node_id] = local_nodes[node_id] + + timers['local_nodes'] = time.time() - start + + # + # XXX Synchronize foreign slice instantiation states + # + + slice_instantiations = SliceInstantiations(self.api).dict() + + # + # Synchronize foreign slices + # + + start = time.time() + + message('(6) Dealing with Slices') + + # Compare only the columns returned by the GetPeerData() call + if peer_tables['Slices']: + columns = peer_tables['Slices'][0].keys() + columns = intersect(columns, Slice.fields) + else: + columns = None + + # Keyed on foreign slice_id + old_peer_slices = Slices( + self.api, {'peer_id': peer_id}, columns).dict('peer_slice_id') + slices_at_peer = dict([(slice['slice_id'], slice) + for slice in peer_tables['Slices']]) + + # Fix up site_id, instantiation, and creator_person_id references + for peer_slice_id, slice in slices_at_peer.items(): + errors = [] + if slice['site_id'] not in peer_sites: + errors.append("invalid site {}".format(slice['site_id'])) + if slice['instantiation'] not in slice_instantiations: + errors.append("invalid instantiation {}" + .format(slice['instantiation'])) + if slice['creator_person_id'] not in peer_persons: + # Just NULL it out + slice['creator_person_id'] = None + else: + slice['creator_person_id'] = peer_persons[ + slice['creator_person_id']]['person_id'] + if errors: + message("Warning: Skipping invalid {peername} slice {slice} : {errors}" + .format(peername=peername, + slice=slice, errors=", ".join(errors))) + del slices_at_peer[peer_slice_id] + continue + else: + slice['site_id'] = peer_sites[slice['site_id']]['site_id'] + + # Synchronize new set + peer_slices = sync(old_peer_slices, slices_at_peer, Slice, + ignore(columns, RefreshPeer.ignore_slice_fields)) + + message('(7) Dealing with Nodes in Slices') + # transcoder : retrieve a local node_id from a peer_node_id + node_transcoder = dict([(node['node_id'], peer_node_id) + for peer_node_id, node in peer_nodes.iteritems()]) + person_transcoder = dict([(person['person_id'], peer_person_id) + for peer_person_id, person in peer_persons.iteritems()]) + + for peer_slice_id, slice in peer_slices.iteritems(): + # Bind any newly cached foreign slices to peer + if peer_slice_id not in old_peer_slices: + peer.add_slice(slice, peer_slice_id, commit=commit_mode) + slice['peer_id'] = peer_id + slice['peer_slice_id'] = peer_slice_id + slice['node_ids'] = [] + slice['person_ids'] = [] + + # Slice as viewed by peer + peer_slice = slices_at_peer[peer_slice_id] + + # Nodes that are currently part of the slice + old_slice_node_ids = [node_transcoder[node_id] for node_id in slice['node_ids'] + if node_id in node_transcoder and node_transcoder[node_id] in peer_nodes] + + # Nodes that should be part of the slice + slice_node_ids = [node_id for node_id in peer_slice['node_ids'] + if node_id in peer_nodes] + + # Remove stale nodes from slice + for node_id in (set(old_slice_node_ids) - set(slice_node_ids)): + slice.remove_node(peer_nodes[node_id], commit=commit_mode) + message("{peername} node {hostname} (id {node_id}) removed from slice {slicename} (id {slice_id})" + .format(peername=peername, + hostname=peer_nodes[node_id]['hostname'], node_id=peer_nodes[node_id]['node_id'], + slicename=slice['name'], slice_id=slice['slice_id'])) + + # Add new nodes to slice + for node_id in (set(slice_node_ids) - set(old_slice_node_ids)): + slice.add_node(peer_nodes[node_id], commit=commit_mode) + message("{peername} node {hostname} (id {node_id}) added into slice {slicename} (id {slice_id})" + .format(peername=peername, + hostname=peer_nodes[node_id]['hostname'], node_id=peer_nodes[node_id]['node_id'], + slicename=slice['name'], slice_id=slice['slice_id'])) + + if slice['slice_id'] == 225: + return + + # N.B.: Local nodes that may have been added to the slice + # by hand, are removed. In other words, don't do this. + + # Foreign users that are currently part of the slice + # old_slice_person_ids = [ person_transcoder[person_id] for person_id in slice['person_ids'] \ + # if person_transcoder[person_id] in peer_persons] + # An issue occurred with a user who registered on both sites (same email) + # So the remote person could not get cached locally + # The one-line map/filter style is nicer but ineffective here + old_slice_person_ids = [] + for person_id in slice['person_ids']: + if not person_transcoder.has_key(person_id): + message('WARNING : person_id {person_id} in {slicename} not transcodable (1) - skipped' + .format(person_id=person_id, slicename=slice['name'])) + elif person_transcoder[person_id] not in peer_persons: + message('WARNING : person_id {person_id} in {slicename} not transcodable (2) - skipped' + .format(person_id=person_id, slicename=slice['name'])) + else: + old_slice_person_ids += [person_transcoder[person_id]] + + # Foreign users that should be part of the slice + slice_person_ids = [person_id for person_id in peer_slice[ + 'person_ids'] if person_id in peer_persons] + + # Remove stale users from slice + for person_id in (set(old_slice_person_ids) - set(slice_person_ids)): + slice.remove_person( + peer_persons[person_id], commit=commit_mode) + message("{peername} user {email} removed from slice {slicename}" + .format(peername=peername, + email=peer_persons[person_id]['email'], + slicename=slice['name'])) + + # Add new users to slice + for person_id in (set(slice_person_ids) - set(old_slice_person_ids)): + slice.add_person(peer_persons[person_id], commit=commit_mode) + message("{peername} user {email} added into slice {slicename}" + .format(peername=peername, + email=peer_persons[person_id]['email'], + slicename=slice['name'])) + + # N.B.: Local users that may have been added to the slice + # by hand, are not touched. + + timers['slices'] = time.time() - start + + # + # Persons x Sites + # + start = time.time() + + message('(8) Dealing with Persons in Sites') + + for peer_site_id, site in peer_sites.iteritems(): + # Site as viewed by peer + peer_site = sites_at_peer[peer_site_id] + + # Persons that are currently part of the site + old_site_person_ids = [person_transcoder[person_id] for person_id in site['person_ids'] + if person_id in person_transcoder and person_transcoder[person_id] in peer_persons] + + # Perons that should be part of the site + site_person_ids = [person_id for person_id in peer_site[ + 'person_ids'] if person_id in peer_persons] + + # Remove stale persons from site + for person_id in (set(old_site_person_ids) - set(site_person_ids)): + site.remove_person(peer_persons[person_id], commit=commit_mode) + message("{peername} person {email} removed from site {login_base}" + .format(peername=peername, + email=peer_persons[person_id]['email'], + login_base=site['login_base'])) + + # Add new persons to site + for person_id in (set(site_person_ids) - set(old_site_person_ids)): + site.add_person(peer_persons[person_id], commit=commit_mode) + message("{peername} person {email} added into site {login_base}" + .format(peername=peername, + email=peer_persons[person_id]['email'], + login_base=site['login_base'])) + + timers['sites-persons'] = time.time() - start + + # + # Persons x Roles + # + start = time.time() + + message('(9) Dealing with Roles for Persons') + + roles = Roles(self.api) + roles_dict = dict([(role['role_id'], role) for role in roles]) + for peer_person_id, person in peer_persons.iteritems(): + # Person as viewed by peer + peer_person = persons_at_peer[peer_person_id] + + # Roles that are currently attributed for the person + old_person_role_ids = [role_id for role_id in person['role_ids']] + + # Roles that should be attributed to the person + person_role_ids = [role_id for role_id in peer_person['role_ids']] + + # Remove stale roles + for role_id in (set(old_person_role_ids) - set(person_role_ids)): + person.remove_role(roles_dict[role_id], commit=commit_mode) + message("{peername} role {rolename} removed from person {email}" + .format(peername=peername, + rolename=roles_dict[role_id]['name'], + email=person['email'])) + + # Add new roles to person + for role_id in (set(person_role_ids) - set(old_person_role_ids)): + person.add_role(roles_dict[role_id], commit=commit_mode) + message("{peername} role {rolename} added from person {email}" + .format(peername=peername, + rolename=roles_dict[role_id]['name'], + email=person['email'])) + + timers['persons-roles'] = time.time() - start + + # Update peer itself and commit + peer.sync(commit=True) + + return timers