# # PlanetLab importer # # requirements # # read the planetlab database and update the local registry database accordingly # (in other words, with this testbed, the SFA registry is *not* authoritative) # so we update the following collections # . authorities (from pl sites) # . node (from pl nodes) # . users+keys (from pl persons and attached keys) # known limitation : *one* of the ssh keys is chosen at random here # xxx todo/check xxx at the very least, when a key is known to the registry # and is still current in plc # then we should definitely make sure to keep that one in sfa... # . slice+researchers (from pl slices and attached users) # import os from sfa.util.config import Config from sfa.util.xrn import Xrn, get_leaf, get_authority, hrn_to_urn from sfa.trust.gid import create_uuid from sfa.trust.certificate import convert_public_key, Keypair from sfa.storage.alchemy import dbsession from sfa.storage.model import RegRecord, RegAuthority, RegSlice, RegNode, RegUser, RegKey from sfa.planetlab.plshell import PlShell from sfa.planetlab.plxrn import hostname_to_hrn, slicename_to_hrn, email_to_hrn, hrn_to_pl_slicename def _get_site_hrn(interface_hrn, site): # Hardcode 'internet2' into the hrn for sites hosting # internet2 nodes. This is a special operation for some vini # sites only hrn = ".".join([interface_hrn, site['login_base']]) if ".vini" in interface_hrn and interface_hrn.endswith('vini'): if site['login_base'].startswith("i2") or site['login_base'].startswith("nlr"): hrn = ".".join([interface_hrn, "internet2", site['login_base']]) return hrn class PlImporter: def __init__ (self, auth_hierarchy, logger): self.auth_hierarchy = auth_hierarchy self.logger=logger def add_options (self, parser): # we don't have any options for now pass # hrn hash is initialized from current db # remember just-created records as we go # xxx might make sense to add a UNIQUE constraint in the db itself def remember_record_by_hrn (self, record): tuple = (record.type, record.hrn) if tuple in self.records_by_type_hrn: self.logger.warning ("PlImporter.remember_record_by_hrn: duplicate (%s,%s)"%tuple) return self.records_by_type_hrn [ tuple ] = record # ditto for pointer hash def remember_record_by_pointer (self, record): if record.pointer == -1: self.logger.warning ("PlImporter.remember_record_by_pointer: pointer is void") return tuple = (record.type, record.pointer) if tuple in self.records_by_type_pointer: self.logger.warning ("PlImporter.remember_record_by_pointer: duplicate (%s,%s)"%tuple) return self.records_by_type_pointer [ ( record.type, record.pointer,) ] = record def remember_record (self, record): self.remember_record_by_hrn (record) self.remember_record_by_pointer (record) def locate_by_type_hrn (self, type, hrn): return self.records_by_type_hrn.get ( (type, hrn), None) def locate_by_type_pointer (self, type, pointer): return self.records_by_type_pointer.get ( (type, pointer), None) # a convenience/helper function to see if a record is already known # a former, broken, attempt (in 2.1-9) had been made # to try and use 'pointer' as a first, most significant attempt # the idea being to preserve stuff as much as possible, and thus # to avoid creating a new gid in the case of a simple hrn rename # however this of course doesn't work as the gid depends on the hrn... #def locate (self, type, hrn=None, pointer=-1): # if pointer!=-1: # attempt = self.locate_by_type_pointer (type, pointer) # if attempt : return attempt # if hrn is not None: # attempt = self.locate_by_type_hrn (type, hrn,) # if attempt : return attempt # return None # this makes the run method a bit abtruse - out of the way def create_special_vini_record (self, interface_hrn): # special case for vini if ".vini" in interface_hrn and interface_hrn.endswith('vini'): # create a fake internet2 site first i2site = {'name': 'Internet2', 'login_base': 'internet2', 'site_id': -1} site_hrn = _get_site_hrn(interface_hrn, i2site) # import if hrn is not in list of existing hrns or if the hrn exists # but its not a site record if ( 'authority', site_hrn, ) not in self.records_by_type_hrn: urn = hrn_to_urn(site_hrn, 'authority') if not self.auth_hierarchy.auth_exists(urn): self.auth_hierarchy.create_auth(urn) auth_info = self.auth_hierarchy.get_auth_info(urn) auth_record = RegAuthority(hrn=site_hrn, gid=auth_info.get_gid_object(), pointer=site['site_id'], authority=get_authority(site_hrn)) auth_record.just_created() dbsession.add(auth_record) dbsession.commit() self.logger.info("PlImporter: Imported authority (vini site) %s"%auth_record) self.remember_record ( site_record ) def run (self, options): config = Config () interface_hrn = config.SFA_INTERFACE_HRN root_auth = config.SFA_REGISTRY_ROOT_AUTH shell = PlShell (config) ######## retrieve all existing SFA objects all_records = dbsession.query(RegRecord).all() # create hash by (type,hrn) # we essentially use this to know if a given record is already known to SFA self.records_by_type_hrn = \ dict ( [ ( (record.type, record.hrn) , record ) for record in all_records ] ) # create hash by (type,pointer) self.records_by_type_pointer = \ dict ( [ ( (record.type, record.pointer) , record ) for record in all_records if record.pointer != -1] ) # initialize record.stale to True by default, then mark stale=False on the ones that are in use for record in all_records: record.stale=True ######## retrieve PLC data # Get all plc sites # retrieve only required stuf sites = shell.GetSites({'peer_id': None, 'enabled' : True}, ['site_id','login_base','node_ids','slice_ids','person_ids',]) # create a hash of sites by login_base # sites_by_login_base = dict ( [ ( site['login_base'], site ) for site in sites ] ) # Get all plc users persons = shell.GetPersons({'peer_id': None, 'enabled': True}, ['person_id', 'email', 'key_ids', 'site_ids', 'role_ids']) # create a hash of persons by person_id persons_by_id = dict ( [ ( person['person_id'], person) for person in persons ] ) # also gather non-enabled user accounts so as to issue relevant warnings disabled_persons = shell.GetPersons({'peer_id': None, 'enabled': False}, ['person_id']) disabled_person_ids = [ person['person_id'] for person in disabled_persons ] # Get all plc public keys # accumulate key ids for keys retrieval key_ids = [] for person in persons: key_ids.extend(person['key_ids']) keys = shell.GetKeys( {'peer_id': None, 'key_id': key_ids, 'key_type': 'ssh'} ) # create a hash of keys by key_id keys_by_id = dict ( [ ( key['key_id'], key ) for key in keys ] ) # create a dict person_id -> [ (plc)keys ] keys_by_person_id = {} for person in persons: pubkeys = [] for key_id in person['key_ids']: # by construction all the keys we fetched are ssh keys # so gpg keys won't be in there try: key = keys_by_id[key_id] pubkeys.append(key) except: self.logger.warning("Could not spot key %d - probably non-ssh"%key_id) keys_by_person_id[person['person_id']] = pubkeys # Get all plc nodes nodes = shell.GetNodes( {'peer_id': None}, ['node_id', 'hostname', 'site_id']) # create hash by node_id nodes_by_id = dict ( [ ( node['node_id'], node, ) for node in nodes ] ) # Get all plc slices slices = shell.GetSlices( {'peer_id': None}, ['slice_id', 'name', 'person_ids']) # create hash by slice_id slices_by_id = dict ( [ (slice['slice_id'], slice ) for slice in slices ] ) # isolate special vini case in separate method self.create_special_vini_record (interface_hrn) # start importing for site in sites: site_hrn = _get_site_hrn(interface_hrn, site) # import if hrn is not in list of existing hrns or if the hrn exists # but its not a site record site_record=self.locate_by_type_hrn ('authority', site_hrn) if not site_record: try: urn = hrn_to_urn(site_hrn, 'authority') if not self.auth_hierarchy.auth_exists(urn): self.auth_hierarchy.create_auth(urn) auth_info = self.auth_hierarchy.get_auth_info(urn) site_record = RegAuthority(hrn=site_hrn, gid=auth_info.get_gid_object(), pointer=site['site_id'], authority=get_authority(site_hrn)) site_record.just_created() dbsession.add(site_record) dbsession.commit() self.logger.info("PlImporter: imported authority (site) : %s" % site_record) self.remember_record (site_record) except: # if the site import fails then there is no point in trying to import the # site's child records (node, slices, persons), so skip them. self.logger.log_exc("PlImporter: failed to import site %s. Skipping child records"%site_hrn) continue else: # xxx update the record ... pass site_record.stale=False # import node records for node_id in site['node_ids']: try: node = nodes_by_id[node_id] except: self.logger.warning ("PlImporter: cannot find node_id %s - ignored"%node_id) continue site_auth = get_authority(site_hrn) site_name = site['login_base'] node_hrn = hostname_to_hrn(site_auth, site_name, node['hostname']) # xxx this sounds suspicious if len(node_hrn) > 64: node_hrn = node_hrn[:64] node_record = self.locate_by_type_hrn ( 'node', node_hrn ) if not node_record: try: pkey = Keypair(create=True) urn = hrn_to_urn(node_hrn, 'node') node_gid = self.auth_hierarchy.create_gid(urn, create_uuid(), pkey) node_record = RegNode (hrn=node_hrn, gid=node_gid, pointer =node['node_id'], authority=get_authority(node_hrn)) node_record.just_created() dbsession.add(node_record) dbsession.commit() self.logger.info("PlImporter: imported node: %s" % node_record) self.remember_record (node_record) except: self.logger.log_exc("PlImporter: failed to import node %s"%node_hrn) continue else: # xxx update the record ... pass node_record.stale=False site_pis=[] # import persons for person_id in site['person_ids']: proceed=False if person_id in persons_by_id: person=persons_by_id[person_id] proceed=True elif person_id in disabled_person_ids: pass else: self.logger.warning ("PlImporter: cannot locate person_id %s in site %s - ignored"%(person_id,site_hrn)) # make sure to NOT run this if anything is wrong if not proceed: continue person_hrn = email_to_hrn(site_hrn, person['email']) # xxx suspicious again if len(person_hrn) > 64: person_hrn = person_hrn[:64] person_urn = hrn_to_urn(person_hrn, 'user') user_record = self.locate_by_type_hrn ( 'user', person_hrn) # return a tuple pubkey (a plc key object) and pkey (a Keypair object) def init_person_key (person, plc_keys): pubkey=None if person['key_ids']: # randomly pick first key in set pubkey = plc_keys[0] try: pkey = convert_public_key(pubkey['key']) except: self.logger.warn('PlImporter: unable to convert public key for %s' % person_hrn) pkey = Keypair(create=True) else: # the user has no keys. Creating a random keypair for the user's gid self.logger.warn("PlImporter: person %s does not have a PL public key"%person_hrn) pkey = Keypair(create=True) return (pubkey, pkey) # new person try: plc_keys = keys_by_person_id.get(person['person_id'],[]) if not user_record: (pubkey,pkey) = init_person_key (person, plc_keys ) person_gid = self.auth_hierarchy.create_gid(person_urn, create_uuid(), pkey, email=person['email']) user_record = RegUser (hrn=person_hrn, gid=person_gid, pointer=person['person_id'], authority=get_authority(person_hrn), email=person['email']) if pubkey: user_record.reg_keys=[RegKey (pubkey['key'], pubkey['key_id'])] else: self.logger.warning("No key found for user %s"%user_record) user_record.just_created() dbsession.add (user_record) dbsession.commit() self.logger.info("PlImporter: imported person: %s" % user_record) self.remember_record ( user_record ) else: # update the record ? # # if a user key has changed then we need to update the # users gid by forcing an update here # # right now, SFA only has *one* key attached to a user, and this is # the key that the GID was made with # so the logic here is, we consider that things are OK (unchanged) if # all the SFA keys are present as PLC keys # otherwise we trigger the creation of a new gid from *some* plc key # and record this on the SFA side # it would make sense to add a feature in PLC so that one could pick a 'primary' # key but this is not available on the myplc side for now # = or = it would be much better to support several keys in SFA but that # does not seem doable without a major overhaul in the data model as # a GID is attached to a hrn, but it's also linked to a key, so... # NOTE: with this logic, the first key entered in PLC remains the one # current in SFA until it is removed from PLC sfa_keys = user_record.reg_keys def sfa_key_in_list (sfa_key,plc_keys): for plc_key in plc_keys: if plc_key['key']==sfa_key.key: return True return False # are all the SFA keys known to PLC ? new_keys=False if not sfa_keys and plc_keys: new_keys=True else: for sfa_key in sfa_keys: if not sfa_key_in_list (sfa_key,plc_keys): new_keys = True if new_keys: (pubkey,pkey) = init_person_key (person, plc_keys) person_gid = self.auth_hierarchy.create_gid(person_urn, create_uuid(), pkey) person_gid.set_email(person['email']) if not pubkey: user_record.reg_keys=[] else: user_record.reg_keys=[ RegKey (pubkey['key'], pubkey['key_id'])] user_record.gid = person_gid user_record.just_updated() self.logger.info("PlImporter: updated person: %s" % user_record) user_record.email = person['email'] dbsession.commit() user_record.stale=False # accumulate PIs - PLCAPI has a limitation that when someone has PI role # this is valid for all sites she is in.. # PI is coded with role_id==20 if 20 in person['role_ids']: site_pis.append (user_record) except: self.logger.log_exc("PlImporter: failed to import person %d %s"%(person['person_id'],person['email'])) # maintain the list of PIs for a given site # for the record, Jordan had proposed the following addition as a welcome hotfix to a previous version: # site_pis = list(set(site_pis)) # this was likely due to a bug in the above logic, that had to do with disabled persons # being improperly handled, and where the whole loop on persons # could be performed twice with the same person... # so hopefully we do not need to eliminate duplicates explicitly here anymore site_record.reg_pis = site_pis dbsession.commit() # import slices for slice_id in site['slice_ids']: try: slice = slices_by_id[slice_id] except: self.logger.warning ("PlImporter: cannot locate slice_id %s - ignored"%slice_id) slice_hrn = slicename_to_hrn(interface_hrn, slice['name']) slice_record = self.locate_by_type_hrn ('slice', slice_hrn) if not slice_record: try: pkey = Keypair(create=True) urn = hrn_to_urn(slice_hrn, 'slice') slice_gid = self.auth_hierarchy.create_gid(urn, create_uuid(), pkey) slice_record = RegSlice (hrn=slice_hrn, gid=slice_gid, pointer=slice['slice_id'], authority=get_authority(slice_hrn)) slice_record.just_created() dbsession.add(slice_record) dbsession.commit() self.logger.info("PlImporter: imported slice: %s" % slice_record) self.remember_record ( slice_record ) except: self.logger.log_exc("PlImporter: failed to import slice %s (%s)"%(slice_hrn,slice['name'])) else: # xxx update the record ... # given that we record the current set of users anyways, there does not seem to be much left to do here # self.logger.warning ("Slice update not yet implemented on slice %s (%s)"%(slice_hrn,slice['name'])) pass # record current users affiliated with the slice slice_record.reg_researchers = \ [ self.locate_by_type_pointer ('user',user_id) for user_id in slice['person_ids'] ] dbsession.commit() slice_record.stale=False ### remove stale records # special records must be preserved system_hrns = [interface_hrn, root_auth, interface_hrn + '.slicemanager'] for record in all_records: if record.hrn in system_hrns: record.stale=False if record.peer_authority: record.stale=False if ".vini" in interface_hrn and interface_hrn.endswith('vini') and \ record.hrn.endswith("internet2"): record.stale=False for record in all_records: try: stale=record.stale except: stale=True self.logger.warning("stale not found with %s"%record) if stale: self.logger.info("PlImporter: deleting stale record: %s" % record) dbsession.delete(record) dbsession.commit()