planetstack/observer/steps/garbage_collector.py

   1 import os
   2 import base64
   3 import traceback
   4 from collections import defaultdict
   5 from django.db.models import F, Q
   6 from planetstack.config import Config
   7 from util.logger import Logger, logging
   8 from observer.openstacksyncstep import OpenStackSyncStep
   9 from core.models import *
  10
  11 logger = Logger(logfile='/var/log/observer.log', level=logging.INFO)
  12
  13 class GarbageCollector(OpenStackSyncStep):
  14     requested_interval = 86400
  15     provides=[]
  16
  17     def call(self, **args):
  18         try:
  19             self.gc_networks()
  20             #self.gc_user_tenant_roles()
  21             #self.gc_tenants()
  22             #self.gc_users()
  23             self.gc_slivers()
  24             #self.gc_sliver_ips()
  25             pass
  26         except:
  27             traceback.print_exc()
  28
  29     def gc_networks(self):
  30         """
  31         Remove all neutron networks that do not exist in the planetstack db.
  32         """
  33         # some networks cannot be deleted
  34         system_networks = ['nat-net','private-admin']
  35         for network_template in NetworkTemplate.objects.all():
  36             if network_template.sharedNetworkName and \
  37               network_template.sharedNetworkName not in system_networks:
  38                 system_networks.append(network_template.sharedNetworkName)
  39
  40         networks = Network.objects.filter(enacted__isnull=False)
  41         networks_dict = {}
  42         for network in networks:
  43             networks_dict[network.name] = network
  44
  45         # some deployments are at the same url. Keep track of the urls we've visited
  46         # to make sure we aren't making redundant calls
  47         completed_urls = []
  48         for deployment in Deployment.objects.all():
  49             # skip deployments that we've already processed
  50             if deployment.auth_url in completed_urls:
  51                 continue
  52             try:
  53                 driver = self.driver.admin_driver(deployment=deployment)
  54                 neutron_networks = driver.shell.quantum.list_networks()['networks']
  55                 for neutron_network in neutron_networks:
  56                     # skip system networks
  57                     if neutron_network['name'] in system_networks:
  58                         continue
  59                     if neutron_network['name'] not in networks_dict:
  60                         try:
  61                             logger.info("GarbageCollector: deleting network %s" % neutron_network['name'])
  62                             for subnet_id in neutron_network['subnets']:
  63                                 driver.delete_subnet(subnet_id)
  64                             driver.delete_network(neutron_network['id'])
  65                         except:
  66                             logger.log_exc("GarbageCollector: delete network %s failed" % neutron_network['name'])
  67             except:
  68                 logger.log_exc("GarbageCollector: Error at deployment %s" % deployment)
  69
  70             completed_urls.append(deployment.auth_url)
  71
  72     def gc_tenants(self):
  73         """
  74         Remove sites and slices that no don't exist in openstack db if they
  75         have an enacted time (enacted != None).
  76         """
  77         # some tenants cannot be deleted
  78         system_tenants = ['admin','service', 'invisible_to_admin']
  79         # get all sites that where enacted != null. We can assume these sites
  80         # have previously been synced and need to be checed for deletion.
  81         sites = Site.objects.filter(enacted__isnull=False)
  82         site_dict = {}
  83         for site in sites:
  84             site_dict[site.login_base] = site
  85
  86         # get all slices that where enacted != null. We can assume these slices
  87         # have previously been synced and need to be checed for deletion.
  88         slices = Slice.objects.filter(enacted__isnull=False)
  89         slice_dict = {}
  90         for slice in slices:
  91             slice_dict[slice.name] = slice
  92
  93         # delete keystone tenants that don't have a site record
  94         # some deployments are at the same url. Keep track of the urls we've visited
  95         # to make sure we aren't making redundant calls
  96         completed_urls = []
  97         for deployment in Deployment.objects.all():
  98             # skip deployments that we've already processed
  99             if deployment.auth_url in completed_urls:
 100                 continue
 101
 102             driver = self.driver.admin_driver(deployment=deployment)
 103             tenants = driver.shell.keystone.tenants.findall()
 104             for tenant in tenants:
 105                 if tenant.name in system_tenants:
 106                     continue
 107                 if tenant.name not in site_dict and tenant.name not in slice_dict:
 108                     try:
 109                         logger.info("GarbageCollector: deleting tenant: %s" % (tenant))
 110                         driver.delete_tenant(tenant.id)
 111                     except:
 112                         logger.log_exc("GarbageCollector: delete tenant failed: %s" % tenant)
 113             completed_urls.append(deployment.auth_url)
 114
 115     def gc_users(self):
 116         """
 117         Remove users that do not exist in openstack db if they have an
 118         enacted time (enacted != None).
 119         """
 120         # some users cannot be deleted
 121         system_users = ['admin', 'nova', 'quantum', 'neutron' 'glance', \
 122                         'cinder', 'swift', 'service', 'demo']
 123
 124         # get all users that where enacted != null. We can assume these users
 125         # have previously been synced and need to be checed for deletion.
 126         users = User.objects.filter(enacted__isnull=False)
 127         user_dict = {}
 128         for user in users:
 129             user_dict[user.kuser_id] = user
 130
 131         # delete keystone users that don't have a user record
 132         # some deployments are at the same url. Keep track of the urls we've visited
 133         # to make sure we aren't making redundant calls
 134         completed_urls = []
 135         for deployment in Deployment.objects.all():
 136             # skip deployments that we've already processed
 137             if deployment.auth_url in completed_urls:
 138                 continue
 139
 140             driver = self.driver.admin_driver(deployment=deployment)
 141             users = driver.shell.keystone.users.findall()
 142             for user in users:
 143                 if user.name in system_users:
 144                     continue
 145                 if user.id not in user_dict:
 146                     try:
 147                         logger.info("GarbageCollector: deleting user: %s" % user)
 148                         self.driver.delete_user(user.id)
 149                     except:
 150                         logger.log_exc("GarbageCollector: delete user failed: %s" % user)
 151             completed_urls.append(deployment.auth_url)
 152
 153     def gc_user_tenant_roles(self):
 154         """
 155         Remove roles that don't exist in openstack db if they have
 156         an enacted time (enacted != None).
 157         """
 158         # get all site privileges and slice memberships that have been enacted
 159         user_tenant_roles = defaultdict(list)
 160         for site_priv in SitePrivilege.objects.filter(enacted__isnull=False):
 161             user_tenant_roles[(site_priv.user.kuser_id, site_priv.site.tenant_id)].append(site_priv.role.role)
 162         for slice_memb in SlicePrivilege.objects.filter(enacted__isnull=False):
 163             user_tenant_roles[(slice_memb.user.kuser_id, slice_memb.slice.tenant_id)].append(slice_memb.role.role)
 164
 165         # some deployments are at the same url. Keep track of the urls we've visited
 166         # to make sure we aren't making redundant calls
 167         completed_urls = []
 168         # Some user tenant role aren't stored in planetstack but they must be preserved.
 169         # Role that fall in this category are
 170         # 1. Never remove a user's role that their home site
 171         # 2. Never remove a user's role at a slice they've created.
 172         # Keep track of all roles that must be preserved.
 173         users = User.objects.all()
 174         for deployment in Deployment.objects.all():
 175             # skip deployments that we've already processed
 176             if deployment.auth_url in completed_urls:
 177                 continue
 178
 179             driver = self.driver.admin_driver(deployment=deployment)
 180             tenants = driver.shell.keystone.tenants.list()
 181             for user in users:
 182                 # skip admin roles
 183                 if user.kuser_id == self.driver.admin_user.id:
 184                     continue
 185
 186                 ignore_tenant_ids = []
 187                 k_user = driver.shell.keystone.users.find(id=user.kuser_id)
 188                 ignore_tenant_ids = [s['tenant_id'] for s in user.slices.values()]
 189                 if user.site:
 190                     ignore_tenant_ids.append(user.site.tenant_id)
 191
 192                 # get user roles in keystone
 193                 for tenant in tenants:
 194                     # skip preserved tenant ids
 195                     if tenant.tenant_id in ignore_tenant_ids:
 196                         continue
 197                     # compare user tenant roles
 198                     user_tenant_role_ids = user_tenant_roles.get((user.kuser_id, tenant.id), [])
 199
 200                     if user_tenant_role_ids:
 201                         # The user has roles at the tenant. Check if roles need to
 202                         # be updated.
 203                         k_user_roles =  driver.shell.keystone.roles.roles_for_user(k_user, tenant)
 204                         for k_user_role in k_user_roles:
 205                             if k_user_role.role_id not in user_tenant_role_ids:
 206                                 logger.info("GarbageCollector: removing user role %s for %s at %s" % \
 207                                            (k_user_role, k_user.username, tenant.name))
 208                                 driver.shell.keyston.remove_user_role(k_user, k_user_role, tenant)
 209                     else:
 210                         # remove all roles the user has at the tenant.
 211                         for k_user_role in k_user_roles:
 212                             logger.info("GarbageCollector: removing user role %s for %s at %s" % \
 213                                        (k_user_role, k_user.username, tenant.name))
 214                             driver.shell.keyston.remove_user_role(k_user, k_user_role, tenant)
 215             completed_urls.append(deployment.auth_url)
 216
 217     def gc_slivers(self):
 218         """
 219         Remove slivers that no don't exist in openstack db if they have
 220         an enacted time (enacted != None).
 221         """
 222         # get all slivers where enacted != null. We can assume these users
 223         # have previously been synced and need to be checed for deletion.
 224         slivers = Sliver.objects.filter(enacted__isnull=False)
 225         sliver_dict = {}
 226         for sliver in slivers:
 227             sliver_dict[sliver.instance_id] = sliver
 228
 229
 230         # some deployments are at the same url. Keep track of the urls we've visited
 231         # to make sure we aren't making redundant calls
 232         completed_urls = []
 233         for deployment in Deployment.objects.all():
 234             # skip deployments that we've already processed
 235             if deployment.auth_url in completed_urls:
 236                 continue
 237
 238             try:
 239                 driver = self.driver.admin_driver(deployment=deployment)
 240                 for tenant in driver.shell.keystone.tenants.list():
 241                     if tenant.name in ['admin', 'services']:
 242                         continue
 243                     # delete sliver that don't have a sliver record
 244                     tenant_driver = self.driver.client_driver(tenant=tenant.name, deployment=deployment)
 245                     for instance in tenant_driver.shell.nova.servers.list():
 246                         if instance.id not in sliver_dict:
 247                             try:
 248                                 logger.info("GarbageCollector: destroying sliver: %s %s" % (instance, instance.id))
 249                                 tenant_driver.destroy_instance(instance.id)
 250                             except:
 251                                 logger.log_exc("GarbageCollector: destroy sliver failed: %s" % instance)
 252             except:
 253                 logger.log_exc("GarbageCollector: Error at deployment %s" % deployment)
 254             completed_urls.append(deployment.auth_url)
 255
 256
 257     def gc_sliver_ips(self):
 258         """
 259         Update ips that have changed.
 260         """
 261         # fill in null ip addresses
 262         slivers = Sliver.objects.filter(ip=None)
 263         for sliver in slivers:
 264             # update connection
 265
 266             driver = self.driver.client_driver(tenant=sliver.slice.name, deployment=sliver.node.deployment)
 267             servers = driver.shell.nova.servers.findall(id=sliver.instance_id)
 268             if not servers:
 269                 continue
 270             server = servers[0]
 271             ips = server.addresses.get(sliver.slice.name, [])
 272             if ips and sliver.ip != ips[0]['addr']:
 273                 sliver.ip = ips[0]['addr']
 274                 sliver.save()
 275                 logger.info("updated sliver ip: %s %s" % (sliver, ips[0]))
 276
 277     def gc_nodes(self):
 278          # collect local nodes
 279         nodes = Node.objects.all()
 280         nodes_dict = {}
 281         for node in nodes:
 282             nodes_dict[node.name] = node
 283
 284         # collect nova nodes:
 285         compute_nodes_dict = {}
 286         for deployment in Deployment.objets.all():
 287             driver = self.driver.admin_driver(deployment=deployment)
 288             compute_nodes = driver.nova.hypervisors.list()
 289             for compute_node in compute_nodes:
 290                 compute_nodes_dict[compute_node.hypervisor_hostname] = compute_node
 291
 292         # remove old nodes
 293         old_node_names = set(nodes_dict.keys()).difference(compute_nodes_dict.keys())
 294         Node.objects.filter(name__in=old_node_names).delete()
 295
 296     def gc_images(self):
 297         # collect local images
 298         images = Image.objects.all()
 299         images_dict = {}
 300         for image in images:
 301             images_dict[image.name] = image
 302
 303         # collect glance images
 304         glance_images_dict = {}
 305         for deployment in Deployment.objects.all():
 306             driver = self.driver.admin_driver(deployment=deployment)
 307             glance_images = driver.shell.glance.get_images()
 308             for glance_image in glance_images:
 309                 glance_images_dict[glance_image['name']] = glance_image
 310
 311         # remove old images
 312         old_image_names = set(images_dict.keys()).difference(glance_images_dict.keys())
 313         Image.objects.filter(name__in=old_image_names).delete()