%PDF- %PDF-
Direktori : /lib/python3/dist-packages/samba/ |
Current File : //lib/python3/dist-packages/samba/uptodateness.py |
# Uptodateness utils # # Copyright (C) Andrew Bartlett 2015, 2018 # Copyright (C) Douglas Bagnall <douglas.bagnall@catalyst.net.nz> # Copyright (C) Joe Guo <joeg@catalyst.net.nz> # # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program. If not, see <http://www.gnu.org/licenses/>. import sys import time from ldb import SCOPE_BASE, LdbError from samba import nttime2unix, dsdb from samba.netcmd import CommandError from samba.samdb import SamDB from samba.kcc import KCC def get_kcc_and_dsas(url, lp, creds): """Get a readonly KCC object and the list of DSAs it knows about.""" unix_now = int(time.time()) kcc = KCC(unix_now, readonly=True) kcc.load_samdb(url, lp, creds) dsa_list = kcc.list_dsas() dsas = set(dsa_list) if len(dsas) != len(dsa_list): print("There seem to be duplicate dsas", file=sys.stderr) return kcc, dsas def get_partition_maps(samdb): """Generate dictionaries mapping short partition names to the appropriate DNs.""" base_dn = samdb.domain_dn() short_to_long = { "DOMAIN": base_dn, "CONFIGURATION": str(samdb.get_config_basedn()), "SCHEMA": "CN=Schema,%s" % samdb.get_config_basedn(), "DNSDOMAIN": "DC=DomainDnsZones,%s" % base_dn, "DNSFOREST": "DC=ForestDnsZones,%s" % base_dn } long_to_short = {} for s, l in short_to_long.items(): long_to_short[l] = s return short_to_long, long_to_short def get_partition(samdb, part): # Allow people to say "--partition=DOMAIN" rather than # "--partition=DC=blah,DC=..." if part is not None: short_partitions, long_partitions = get_partition_maps(samdb) part = short_partitions.get(part.upper(), part) if part not in long_partitions: raise CommandError("unknown partition %s" % part) return part def get_utdv(samdb, dn): """This finds the uptodateness vector in the database.""" cursors = [] config_dn = samdb.get_config_basedn() for c in dsdb._dsdb_load_udv_v2(samdb, dn): inv_id = str(c.source_dsa_invocation_id) res = samdb.search(base=config_dn, expression=("(&(invocationId=%s)" "(objectClass=nTDSDSA))" % inv_id), attrs=["distinguishedName", "invocationId"]) try: settings_dn = str(res[0]["distinguishedName"][0]) prefix, dsa_dn = settings_dn.split(',', 1) except IndexError as e: continue if prefix != 'CN=NTDS Settings': raise CommandError("Expected NTDS Settings DN, got %s" % settings_dn) cursors.append((dsa_dn, inv_id, int(c.highest_usn), nttime2unix(c.last_sync_success))) return cursors def get_own_cursor(samdb): res = samdb.search(base="", scope=SCOPE_BASE, attrs=["highestCommittedUSN"]) usn = int(res[0]["highestCommittedUSN"][0]) now = int(time.time()) return (usn, now) def get_utdv_edges(local_kcc, dsas, part_dn, lp, creds): # we talk to each remote and make a matrix of the vectors # for each partition # normalise by oldest utdv_edges = {} for dsa_dn in dsas: res = local_kcc.samdb.search(dsa_dn, scope=SCOPE_BASE, attrs=["dNSHostName"]) ldap_url = "ldap://%s" % res[0]["dNSHostName"][0] try: samdb = SamDB(url=ldap_url, credentials=creds, lp=lp) cursors = get_utdv(samdb, part_dn) own_usn, own_time = get_own_cursor(samdb) remotes = {dsa_dn: own_usn} for dn, guid, usn, t in cursors: remotes[dn] = usn except LdbError as e: print("Could not contact %s (%s)" % (ldap_url, e), file=sys.stderr) continue utdv_edges[dsa_dn] = remotes return utdv_edges def get_utdv_distances(utdv_edges, dsas): distances = {} for dn1 in dsas: try: peak = utdv_edges[dn1][dn1] except KeyError as e: peak = 0 d = {} distances[dn1] = d for dn2 in dsas: if dn2 in utdv_edges: if dn1 in utdv_edges[dn2]: dist = peak - utdv_edges[dn2][dn1] d[dn2] = dist else: print(f"Missing dn {dn1} from UTD vector for dsa {dn2}", file=sys.stderr) else: print("missing dn %s from UTD vector list" % dn2, file=sys.stderr) return distances def get_utdv_max_distance(distances): max_distance = 0 for vector in distances.values(): for distance in vector.values(): max_distance = max(max_distance, distance) return max_distance def get_utdv_summary(distances, filters=None): maximum = failure = 0 median = 0.0 # could be average of 2 median values values = [] # put all values into a list, exclude self to self ones for dn_outer, vector in distances.items(): for dn_inner, distance in vector.items(): if dn_outer != dn_inner: values.append(distance) if values: values.sort() maximum = values[-1] length = len(values) if length % 2 == 0: index = length//2 - 1 median = (values[index] + values[index+1])/2.0 median = round(median, 1) # keep only 1 decimal digit like 2.5 else: index = (length - 1)//2 median = values[index] median = float(median) # ensure median is always a float like 1.0 # if value not exist, that's a failure expected_length = len(distances) * (len(distances) - 1) failure = expected_length - length summary = { 'maximum': maximum, 'median': median, 'failure': failure, } if filters: return {key: summary[key] for key in filters} else: return summary