planetstack/openstack/observer.py

   1 import time
   2 import traceback
   3 import commands
   4 import threading
   5
   6 from datetime import datetime
   7 from collections import defaultdict
   8 from core.models import *
   9 from django.db.models import F, Q
  10 from openstack.manager import OpenStackManager
  11 from util.logger import Logger, logging, logger
  12 #from timeout import timeout
  13
  14
  15 logger = Logger(logfile='observer.log', level=logging.INFO)
  16
  17 class OpenStackObserver:
  18
  19     def __init__(self):
  20         self.manager = OpenStackManager()
  21         # The Condition object that gets signalled by Feefie events
  22         self.event_cond = threading.Condition()
  23
  24     def wait_for_event(self, timeout):
  25         self.event_cond.acquire()
  26         self.event_cond.wait(timeout)
  27         self.event_cond.release()
  28
  29     def wake_up(self):
  30         logger.info('Wake up routine called. Event cond %r'%self.event_cond)
  31         self.event_cond.acquire()
  32         self.event_cond.notify()
  33         self.event_cond.release()
  34
  35     def run(self):
  36         if not self.manager.enabled or not self.manager.has_openstack:
  37             return
  38         while True:
  39             try:
  40                 logger.info('Observer run loop')
  41                 #self.sync_roles()
  42
  43                 logger.info('Calling sync tenants')
  44                 try:
  45                     self.sync_tenants()
  46                 except:
  47                     logger.log_exc("Exception in sync_tenants")
  48                     traceback.print_exc()
  49
  50                 logger.info('Calling sync users')
  51                 try:
  52                     self.sync_users()
  53                 except:
  54                     logger.log_exc("Exception in sync_users")
  55                     traceback.print_exc()
  56
  57                 logger.info('Calling sync tenant roles')
  58                 try:
  59                     self.sync_user_tenant_roles()
  60                 except:
  61                     logger.log_exc("Exception in sync_users")
  62                     traceback.print_exc()
  63
  64                 logger.info('Calling sync slivers')
  65                 try:
  66                     self.sync_slivers()
  67                 except:
  68                     logger.log_exc("Exception in sync slivers")
  69                     traceback.print_exc()
  70
  71                 logger.info('Calling sync sliver ips')
  72                 try:
  73                     self.sync_sliver_ips()
  74                 except:
  75                     logger.log_exc("Exception in sync_sliver_ips")
  76                     traceback.print_exc()
  77
  78                 logger.info('Calling sync networks')
  79                 try:
  80                     self.sync_networks()
  81                 except:
  82                     logger.log_exc("Exception in sync_networks")
  83                     traceback.print_exc()
  84
  85                 logger.info('Calling sync network slivers')
  86                 try:
  87                     self.sync_network_slivers()
  88                 except:
  89                     logger.log_exc("Exception in sync_network_slivers")
  90                     traceback.print_exc()
  91
  92                 logger.info('Calling sync external routes')
  93                 try:
  94                     self.sync_external_routes()
  95                 except:
  96                      logger.log_exc("Exception in sync_external_routes")
  97                      traceback.print_exc()
  98
  99                 logger.info('Waiting for event')
 100                 tBeforeWait = time.time()
 101                 self.wait_for_event(timeout=300)
 102
 103                 # Enforce 5 minutes between wakeups
 104                 tSleep = 300 - (time.time() - tBeforeWait)
 105                 if tSleep > 0:
 106                     logger.info('Sleeping for %d seconds' % tSleep)
 107                     time.sleep(tSleep)
 108
 109                 logger.info('Observer woken up')
 110             except:
 111                 logger.log_exc("Exception in observer run loop")
 112                 traceback.print_exc()
 113
 114     def sync_roles(self):
 115         """
 116         save all role that don't already exist in keystone. Remove keystone roles that
 117         don't exist in planetstack
 118         """
 119         # sync all roles that don't already in keystone
 120         keystone_roles = self.manager.driver.shell.keystone.roles.findall()
 121         keystone_role_names = [kr.name for kr in keystone_roles]
 122         pending_roles = Role.objects.all()
 123         pending_role_names = [r.role_type for r in pending_roles]
 124         for role in pending_roles:
 125             if role.role_type not in keystone_role_names:
 126                 try:
 127                     self.manager.save_role(role)
 128                     logger.info("save role: %s" % (role))
 129                 except:
 130                     logger.log_exc("save role failed: %s" % role)
 131                     traceback.print_exc()
 132
 133         # don't delete roles for now
 134         """
 135         # delete keystone roles that don't exist in planetstack
 136         for keystone_role in keystone_roles:
 137             if keystone_role.name == 'admin':
 138                 continue
 139             if keystone_role.name not in pending_role_names:
 140                 try:
 141                     self.manager.driver.delete_role({id: keystone_role.id})
 142                 except:
 143                     traceback.print_exc()
 144         """
 145
 146     def sync_tenants(self):
 147         """
 148         Save all sites and sliceswhere enacted < updated or enacted == None.
 149         Remove sites and slices that no don't exist in openstack db if they
 150         have an enacted time (enacted != None).
 151         """
 152         # get all sites that need to be synced (enacted < updated or enacted is None)
 153         pending_sites = Site.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 154         for site in pending_sites:
 155             try:
 156                 self.manager.save_site(site)
 157                 logger.info("saved site %s" % site)
 158             except:
 159                 logger.log_exc("save site failed: %s" % site)
 160
 161         # get all slices that need to be synced (enacted < updated or enacted is None)
 162         pending_slices = Slice.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 163         for slice in pending_slices:
 164             try:
 165                 self.manager.init_caller(slice.creator, slice.creator.site.login_base)
 166                 self.manager.save_slice(slice)
 167                 logger.info("saved slice %s" % slice)
 168             except:
 169                 logger.log_exc("save slice failed: %s" % slice)
 170
 171         # get all sites that where enacted != null. We can assume these sites
 172         # have previously been synced and need to be checed for deletion.
 173         sites = Site.objects.filter(enacted__isnull=False)
 174         site_dict = {}
 175         for site in sites:
 176             site_dict[site.login_base] = site
 177
 178         # get all slices that where enacted != null. We can assume these slices
 179         # have previously been synced and need to be checed for deletion.
 180         slices = Slice.objects.filter(enacted__isnull=False)
 181         slice_dict = {}
 182         for slice in slices:
 183             slice_dict[slice.name] = slice
 184
 185         # delete keystone tenants that don't have a site record
 186         tenants = self.manager.driver.shell.keystone.tenants.findall()
 187         system_tenants = ['admin','service']
 188         for tenant in tenants:
 189             if tenant.name in system_tenants:
 190                 continue
 191             if tenant.name not in site_dict and tenant.name not in slice_dict:
 192                 try:
 193                     self.manager.driver.delete_tenant(tenant.id)
 194                     logger.info("deleted tenant: %s" % (tenant))
 195                 except:
 196                     logger.log_exc("delete tenant failed: %s" % tenant)
 197
 198
 199     def sync_users(self):
 200         """
 201         save all users where enacted < updated or enacted == None. Remove users that
 202         no don't exist in openstack db if they have an enacted time (enacted != None).
 203         """
 204         # get all users that need to be synced (enacted < updated or enacted is None)
 205         pending_users = User.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 206         for user in pending_users:
 207             try:
 208                 self.manager.save_user(user)
 209                 logger.info("saved user: %s" % (user))
 210             except:
 211                 logger.log_exc("save user failed: %s" %user)
 212
 213         # get all users that where enacted != null. We can assume these users
 214         # have previously been synced and need to be checed for deletion.
 215         users = User.objects.filter(enacted__isnull=False)
 216         user_dict = {}
 217         for user in users:
 218             user_dict[user.kuser_id] = user
 219
 220         # delete keystone users that don't have a user record
 221         system_users = ['admin', 'nova', 'quantum', 'glance', 'cinder', 'swift', 'service']
 222         users = self.manager.driver.shell.keystone.users.findall()
 223         for user in users:
 224             if user.name in system_users:
 225                 continue
 226             if user.id not in user_dict:
 227                 try:
 228                     #self.manager.driver.delete_user(user.id)
 229                     logger.info("deleted user: %s" % user)
 230                 except:
 231                     logger.log_exc("delete user failed: %s" % user)
 232
 233
 234     def sync_user_tenant_roles(self):
 235         """
 236         Save all site privileges and slice memberships wheree enacted < updated or
 237         enacted == None. Remove ones that don't exist in openstack db if they have
 238         an enacted time (enacted != None).
 239         """
 240         # sync site privileges
 241         pending_site_privileges = SitePrivilege.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 242         for site_priv in pending_site_privileges:
 243             try:
 244                 self.manager.save_site_privilege(site_priv)
 245                 logger.info("saved site privilege: %s" % (site_priv))
 246             except: logger.log_exc("save site privilege failed: %s " % site_priv)
 247
 248         # sync slice memberships
 249         pending_slice_memberships = SliceMembership.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 250         for slice_memb in pending_slice_memberships:
 251             try:
 252                 self.manager.save_slice_membership(slice_memb)
 253                 logger.info("saved slice membership: %s" % (slice_memb))
 254             except: logger.log_exc("save slice membership failed: %s" % slice_memb)
 255
 256         # get all site privileges and slice memberships that have been enacted
 257         user_tenant_roles = defaultdict(list)
 258         for site_priv in SitePrivilege.objects.filter(enacted__isnull=False):
 259             user_tenant_roles[(site_priv.user.kuser_id, site_priv.site.tenant_id)].append(site_priv.role.role)
 260         for slice_memb in SliceMembership.objects.filter(enacted__isnull=False):
 261             user_tenant_roles[(slice_memb.user.kuser_id, slice_memb.slice.tenant_id)].append(slice_memb.role.role)
 262
 263         # Some user tenant role aren't stored in planetstack but they must be preserved.
 264         # Role that fall in this category are
 265         # 1. Never remove a user's role that their home site
 266         # 2. Never remove a user's role at a slice they've created.
 267         # Keep track of all roles that must be preserved.
 268         users = User.objects.all()
 269         preserved_roles = {}
 270         for user in users:
 271             tenant_ids = [s['tenant_id'] for s in user.slices.values()]
 272             tenant_ids.append(user.site.tenant_id)
 273             preserved_roles[user.kuser_id] = tenant_ids
 274
 275
 276         # begin removing user tenant roles from keystone. This is stored in the
 277         # Metadata table.
 278         for metadata in self.manager.driver.shell.keystone_db.get_metadata():
 279             # skip admin roles
 280             if metadata.user_id == self.manager.driver.admin_user.id:
 281                 continue
 282             # skip preserved tenant ids
 283             if metadata.user_id in preserved_roles and \
 284                metadata.tenant_id in preserved_roles[metadata.user_id]:
 285                 continue
 286             # get roles for user at this tenant
 287             user_tenant_role_ids = user_tenant_roles.get((metadata.user_id, metadata.tenant_id), [])
 288
 289             if user_tenant_role_ids:
 290                 # The user has roles at the tenant. Check if roles need to
 291                 # be updated.
 292                 user_keystone_role_ids = metadata.data.get('roles', [])
 293                 for role_id in user_keystone_role_ids:
 294                     if role_id not in user_tenant_role_ids:
 295                         user_keystone_role_ids.pop(user_keystone_role_ids.index(role_id))
 296             else:
 297                 # The user has no roles at this tenant.
 298                 metadata.data['roles'] = []
 299             #session.add(metadata)
 300             logger.info("pruning metadata for %s at %s" % (metadata.user_id, metadata.tenant_id))
 301
 302     def sync_slivers(self):
 303         """
 304         save all slivers where enacted < updated or enacted == None. Remove slivers that
 305         no don't exist in openstack db if they have an enacted time (enacted != None).
 306         """
 307         # get all users that need to be synced (enacted < updated or enacted is None)
 308         pending_slivers = Sliver.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 309         for sliver in pending_slivers:
 310             if sliver.creator:
 311                 try:
 312                     # update manager context
 313                     self.manager.init_caller(sliver.creator, sliver.slice.name)
 314                     self.manager.save_sliver(sliver)
 315                     logger.info("saved sliver: %s" % (sliver))
 316                 except:
 317                     logger.log_exc("save sliver failed: %s" % sliver)
 318
 319         # get all slivers where enacted != null. We can assume these users
 320         # have previously been synced and need to be checed for deletion.
 321         slivers = Sliver.objects.filter(enacted__isnull=False)
 322         sliver_dict = {}
 323         for sliver in slivers:
 324             sliver_dict[sliver.instance_id] = sliver
 325
 326         # delete sliver that don't have a sliver record
 327         ctx = self.manager.driver.shell.nova_db.ctx
 328         instances = self.manager.driver.shell.nova_db.instance_get_all(ctx)
 329         for instance in instances:
 330             if instance.uuid not in sliver_dict:
 331                 try:
 332                     # lookup tenant and update context
 333                     tenant = self.manager.driver.shell.keystone.tenants.find(id=instance.project_id)
 334                     self.manager.init_admin(tenant=tenant.name)
 335                     self.manager.driver.destroy_instance(instance.uuid)
 336                     logger.info("destroyed sliver: %s" % (instance))
 337                 except:
 338                     logger.log_exc("destroy sliver failed: %s" % instance)
 339
 340
 341     def sync_sliver_ips(self):
 342         # fill in null ip addresses
 343         slivers = Sliver.objects.filter(ip=None)
 344         for sliver in slivers:
 345             # update connection
 346             self.manager.init_admin(tenant=sliver.slice.name)
 347             servers = self.manager.driver.shell.nova.servers.findall(id=sliver.instance_id)
 348             if not servers:
 349                 continue
 350             server = servers[0]
 351             ips = server.addresses.get(sliver.slice.name, [])
 352             if not ips:
 353                 continue
 354             sliver.ip = ips[0]['addr']
 355             sliver.save()
 356             logger.info("saved sliver ip: %s %s" % (sliver, ips[0]))
 357
 358     def sync_external_routes(self):
 359         routes = self.manager.driver.get_external_routes()
 360         subnets = self.manager.driver.shell.quantum.list_subnets()['subnets']
 361         for subnet in subnets:
 362             try:
 363                 self.manager.driver.add_external_route(subnet, routes)
 364             except:
 365                 logger.log_exc("failed to add external route for subnet %s" % subnet)
 366
 367     def sync_network_slivers(self):
 368         networkSlivers = NetworkSliver.objects.all()
 369         networkSlivers_by_id = {}
 370         networkSlivers_by_port = {}
 371         for networkSliver in networkSlivers:
 372             networkSlivers_by_id[networkSliver.id] = networkSliver
 373             networkSlivers_by_port[networkSliver.port_id] = networkSliver
 374
 375         networks = Network.objects.all()
 376         networks_by_id = {}
 377         for network in networks:
 378             networks_by_id[network.network_id] = network
 379
 380         slivers = Sliver.objects.all()
 381         slivers_by_instance_id = {}
 382         for sliver in slivers:
 383             slivers_by_instance_id[sliver.instance_id] = sliver
 384
 385         ports = self.manager.driver.shell.quantum.list_ports()["ports"]
 386         for port in ports:
 387             if port["id"] in networkSlivers_by_port:
 388                 # we already have it
 389                 print "already accounted for port", port["id"]
 390                 continue
 391
 392             if port["device_owner"] != "compute:nova":
 393                 # we only want the ports that connect to instances
 394                 continue
 395
 396             network = networks_by_id.get(port['network_id'], None)
 397             if not network:
 398                 #print "no network for port", port["id"], "network", port["network_id"]
 399                 continue
 400
 401             sliver = slivers_by_instance_id.get(port['device_id'], None)
 402             if not sliver:
 403                 print "no sliver for port", port["id"], "device_id", port['device_id']
 404                 continue
 405
 406             if network.template.sharedNetworkId is not None:
 407                 # If it's a shared network template, then more than one network
 408                 # object maps to the quantum network. We have to do a whole bunch
 409                 # of extra work to find the right one.
 410                 networks = network.template.network_set.all()
 411                 network = None
 412                 for candidate_network in networks:
 413                     if (candidate_network.owner == sliver.slice):
 414                         print "found network", candidate_network
 415                         network = candidate_network
 416
 417                 if not network:
 418                     print "failed to find the correct network for a shared template for port", port["id"], "network", port["network_id"]
 419                     continue
 420
 421             if not port["fixed_ips"]:
 422                 print "port", port["id"], "has no fixed_ips"
 423                 continue
 424
 425 #            print "XXX", port
 426
 427             ns = NetworkSliver(network=network,
 428                                sliver=sliver,
 429                                ip=port["fixed_ips"][0]["ip_address"],
 430                                port_id=port["id"])
 431             ns.save()
 432
 433     def sync_networks(self):
 434         """
 435         save all networks where enacted < updated or enacted == None. Remove networks that
 436         no don't exist in openstack db if they have an enacted time (enacted != None).
 437         """
 438         # get all users that need to be synced (enacted < updated or enacted is None)
 439         pending_networks = Network.objects.filter(Q(enacted__lt=F('updated')) | Q(enacted=None))
 440         for network in pending_networks:
 441             if network.owner and network.owner.creator:
 442                 try:
 443                     # update manager context
 444                     self.manager.init_caller(network.owner.creator, network.owner.name)
 445                     self.manager.save_network(network)
 446                     logger.info("saved network: %s" % (network))
 447                 except:
 448                     logger.log_exc("save network failed: %s" % network)
 449
 450         # get all networks where enacted != null. We can assume these users
 451         # have previously been synced and need to be checed for deletion.
 452         networks = Network.objects.filter(enacted__isnull=False)
 453         network_dict = {}
 454         for network in networks:
 455             network_dict[network.network_id] = network
 456
 457         # TODO: delete Network objects if quantum network doesn't exist
 458         #       (need to write self.manager.driver.shell.quantum_db)
 459