def get(self, schedule_id):
"""
Get the call request and the schedule for the given schedule id
@param schedule_id: id of the schedule for the call request
@type schedule_id: str
@return: scheduled call report dictionary
@rtype: dict
"""
if isinstance(schedule_id, basestring):
schedule_id = ObjectId(schedule_id)
scheduled_call = self.scheduled_call_collection.find_one({'_id': schedule_id})
if scheduled_call is None:
raise pulp_exceptions.MissingResource(schedule=str(schedule_id))
report = scheduled_call_to_report_dict(scheduled_call)
return report
def test_expired_object_id(self):
"""
Make sure that _create_expired_object_id() generates correct ObjectIds.
"""
expired_oid = _create_expired_object_id(timedelta(seconds=1))
# The oid should be about a second, but since we didn't create the now_oid until after we
# ran the _create_expired_object_id() function, we can't assert that the timedelta is
# exactly one second. It should definitely be less than two seconds of difference between
# them, however (unless Nose is really struggling to run the tests), so we can make sure
# it fits inside the window of 1 - 2 seconds.
now_oid = ObjectId()
self.assertTrue((now_oid.generation_time -
expired_oid.generation_time) >= timedelta(seconds=1))
self.assertTrue((now_oid.generation_time -
expired_oid.generation_time) < timedelta(seconds=2))
# Also, let's make sure the type is correct
self.assertTrue(isinstance(expired_oid, ObjectId))
def test_update_only_config(self):
# Setup
manager = manager_factory.event_listener_manager()
created = manager.create(http.TYPE_ID, {'a' : 'a', 'b' : 'b'}, [event_data.TYPE_REPO_SYNC_STARTED])
# Test
new_config = {'a' : 'x', 'c' : 'c'}
body = {
'notifier_config' : new_config,
}
status, body = self.put('/v2/events/%s/' % created['id'], body)
# Verify
self.assertEqual(200, status)
updated = EventListener.get_collection().find_one({'_id' : ObjectId(created['_id'])})
expected_config = {'a' : 'x', 'b' : 'b', 'c' : 'c'}
self.assertEqual(updated['notifier_config'], expected_config)
def _repos(v1_database, v2_database):
v1_coll = v1_database.repos
v2_coll = v2_database.repos
# Idempotency: By repo_id
v2_repo_ids = [x['id'] for x in v2_coll.find({}, {'id': 1})]
spec = {
'$and': [
{
'id': {
'$nin': v2_repo_ids
}
},
{
'content_types': V1_ISO_REPO
},
]
}
missing_v1_repos = v1_coll.find(spec)
new_repos = []
for v1_repo in missing_v1_repos:
id = ObjectId()
# Identifying tag for the CLI
v2_notes = v1_repo.get('notes', {})
v2_notes[REPO_NOTE_KEY] = REPO_NOTE_ISO
v2_repo = {
'_id': id, # technically not needed but added for clarity
'id': v1_repo['id'],
'display_name': v1_repo['name'],
'description': None,
'notes': v2_notes,
'scratchpad': {},
'content_unit_count': 0
}
new_repos.append(v2_repo)
if new_repos:
v2_coll.insert(new_repos, safe=True)
return True
def update(self, schedule_id, **schedule_updates):
"""
Update a scheduled call request
Valid schedule updates:
* call_request
* schedule
* failure_threshold
* remaining_runs
* enabled
@param schedule_id: id of the schedule for the call request
@type schedule_id: str
@param schedule_updates: updates for scheduled call
@type schedule_updates: dict
"""
if isinstance(schedule_id, basestring):
schedule_id = ObjectId(schedule_id)
if self.scheduled_call_collection.find_one(schedule_id) is None:
raise pulp_exceptions.MissingResource(schedule=str(schedule_id))
validate_schedule_updates(schedule_updates)
call_request = schedule_updates.pop('call_request', None)
if call_request is not None:
schedule_updates['serialized_call_request'] = call_request.serialize()
schedule = schedule_updates.get('schedule', None)
if schedule is not None:
interval, start, runs = dateutils.parse_iso8601_interval(schedule)
schedule_updates.setdefault('remaining_runs', runs) # honor explicit update
# XXX (jconnor) it'd be nice to update the next_run if the schedule
# has changed, but it requires mucking with the internals of the
# of the scheduled call instance, which is all encapsulated in the
# ScheduledCall constructor
# the next_run field will be correctly updated after the next run
self.scheduled_call_collection.update({'_id': schedule_id}, {'$set': schedule_updates}, safe=True)
def _associate_distribution(v1_distribution, v2_distribution, v2_ass_coll):
# This functions just like errata associations so check _errata for comments
# on this approach.
new_associations = []
for repo_id in v1_distribution['repoids']:
new_association = {
'_id': ObjectId(),
'repo_id': repo_id,
'unit_id': v2_distribution['_id'],
'unit_type_id': 'distribution',
'owner_type': DEFAULT_OWNER_TYPE,
'owner_id': DEFAULT_OWNER_ID,
'created': DEFAULT_CREATED,
'updated': DEFAULT_UPDATED,
}
new_associations.append(new_association)
if new_associations:
v2_ass_coll.insert(new_associations, safe=True)
def _consumer_bindings(v2_database, v1_consumer):
v2_coll = v2_database.consumer_bindings
consumer_id = v1_consumer['id']
repo_ids = v1_consumer['repoids']
# Idempotency: Uniqueness is determined by the tuple of consumer and repo ID
bound_repo_ids = [x['repo_id'] for x in v2_coll.find({'consumer_id' : consumer_id})]
unbound_repo_ids = set(repo_ids) - set(bound_repo_ids)
new_bindings = []
for repo_id in unbound_repo_ids:
binding = {
'_id' : ObjectId(),
'consumer_id' : consumer_id,
'repo_id' : repo_id,
'distributor_id' : YUM_DISTRIBUTOR_ID,
'deleted' : False,
}
new_bindings.append(binding)
if new_bindings:
v2_coll.insert(new_bindings, safe=True)
def _consumer_groups(v1_database, v2_database):
v1_coll = v1_database.consumergroups # seriously, no underscore
v2_coll = v2_database.consumer_groups
# Idempotency: By consumer group ID
v1_group_ids = [x['id'] for x in v1_coll.find({}, {'id' : 1})]
v2_existing_ids = [x['id'] for x in v2_coll.find({}, {'id' : 1})]
ids_to_add = set(v1_group_ids) - set(v2_existing_ids)
for group_id in ids_to_add:
v1_group = v1_coll.find_one({'id' : group_id})
v2_group = {
'_id' : ObjectId(),
'id' : v1_group['id'],
'display_name' : v1_group['id'], # no display name in v1
'description' : v1_group['description'],
'consumer_ids' : v1_group['consumerids'], # again, no underscores
'notes' : v1_group['key_value_pairs'],
'scratchpad' : None, # no idea why this is in the model, this doesn't make any sense
}
# Likely a small amount of groups, not worried about batching the inserts
v2_coll.insert(v2_group, safe=True)
def _package_group_categories(v1_database, v2_database, report):
# These act nearly identically to groups, so see the comments in there
# for more information.
# Idempotency: As with groups, pre-load the tuples of repo ID to category
# ID into memory and use that to check for the category's existed before
# inserting.
v2_coll = v2_database.units_package_category
v2_ass_coll = v2_database.repo_content_units
# Tuple of repo ID and group ID
already_added_tuples = [(x['repo_id'], x['id'])
for x in v2_coll.find({}, {
'repo_id': 1,
'id': 1
})]
v1_repos = v1_database.repos.find({}, {
'id': 1,
'packagegroupcategories': 1
})
for v1_repo in v1_repos:
for category_id in v1_repo.get('packagegroupcategories', {}).keys():
# Idempotency check
if (v1_repo['id'], category_id) in already_added_tuples:
continue
v1_category = v1_repo['packagegroupcategories'][category_id]
category_id = str(uuid.uuid4())
new_category = {
'_id': category_id,
'_storage_path': None,
'_content_type_id': 'package_category',
'description': v1_category['description'],
'display_order': v1_category['display_order'],
'id': v1_category['id'],
'name': v1_category['name'],
'packagegroupids': v1_category['packagegroupids'],
'repo_id': v1_repo['id'],
'translated_description':
v1_category['translated_description'],
'translated_name': v1_category['translated_name'],
}
v2_coll.insert(new_category, safe=True)
new_association = {
'_id': ObjectId(),
'repo_id': v1_repo['id'],
'unit_id': category_id,
'unit_type_id': 'package_category',
'owner_type': DEFAULT_OWNER_TYPE,
'owner_id': DEFAULT_OWNER_ID,
'created': DEFAULT_CREATED,
'updated': DEFAULT_UPDATED,
}
v2_ass_coll.insert(new_association, safe=True)
return True
def _package_groups(v1_database, v2_database, report):
# In v2, the unique identifier for a package group is the
# pairing of the group ID and the repository it's in. In v1, the package
# group is embedded in the repo document itself, which is where we get
# that information from.
# In v1 the repository owns the relationship to a package group. Don't look
# at the model class itself, it wasn't added there, but the code will still
# stuff it in under the key "packagegroups". The value at that key is a dict
# of group ID to a PackageGroup instance (v1 model).
# Idempotency: The simplest way to handle this is to pre-load the set of
# repo ID/group ID tuples into memory and verify each group found in each
# v1 repo against that to determine if it has successfully been added or
# not. The nice part about this over letting the uniqueness checks in mongo
# itself is that we can batch the group inserts.
v2_coll = v2_database.units_package_group
v2_ass_coll = v2_database.repo_content_units
# Tuple of repo ID and group ID
already_added_tuples = [(x['repo_id'], x['id'])
for x in v2_coll.find({}, {
'repo_id': 1,
'id': 1
})]
v1_repos = v1_database.repos.find({}, {'id': 1, 'packagegroups': 1})
for v1_repo in v1_repos:
for group_id in v1_repo.get('packagegroups', {}).keys():
# Idempotency check
if (v1_repo['id'], group_id) in already_added_tuples:
continue
v1_group = v1_repo['packagegroups'][group_id]
v2_group_id = str(uuid.uuid4())
new_group = {
'_id': v2_group_id,
'_storage_path': None,
'_content_type_id': 'package_group',
'conditional_package_names':
v1_group['conditional_package_names'],
'default': v1_group['default'],
'default_package_names': v1_group['default_package_names'],
'description': v1_group['description'],
'display_order': v1_group['display_order'],
'id': v1_group['id'],
'langonly': v1_group['langonly'],
'mandatory_package_names': v1_group['mandatory_package_names'],
'name': v1_group['name'],
'optional_package_names': v1_group['optional_package_names'],
'repo_id': v1_repo['id'],
'translated_description': v1_group['translated_description'],
'translated_name': v1_group['translated_name'],
'user_visible': v1_group['user_visible'],
}
v2_coll.insert(new_group, safe=True)
new_association = {
'_id': ObjectId(),
'repo_id': v1_repo['id'],
'unit_id': v2_group_id,
'unit_type_id': 'package_group',
'owner_type': DEFAULT_OWNER_TYPE,
'owner_id': DEFAULT_OWNER_ID,
'created': DEFAULT_CREATED,
'updated': DEFAULT_UPDATED,
}
v2_ass_coll.insert(new_association, safe=True)
return True
def _errata(v1_database, v2_database, report):
v1_coll = v1_database.errata
v2_coll = v2_database.units_erratum
v2_ass_coll = v2_database.repo_content_units
# Idempotency: We're lucky here, the uniqueness is just by ID, so we can
# do a pre-fetch and determine what needs to be added.
v2_errata_ids = [x['id'] for x in v2_coll.find({}, {'id': 1})]
missing_v1_errata = v1_coll.find({'id': {'$nin': v2_errata_ids}})
for v1_erratum in missing_v1_errata:
erratum_id = str(uuid.uuid4())
new_erratum = {
'_id': erratum_id,
'_storage_path': None,
'_content_type_id': 'erratum',
'description': v1_erratum['description'],
'from_str': v1_erratum['from_str'],
'id': v1_erratum['id'],
'issued': v1_erratum['issued'],
'pkglist': v1_erratum.get('pkglist', []),
'pushcount': v1_erratum['pushcount'],
'reboot_suggested': v1_erratum['reboot_suggested'],
'references': v1_erratum['references'],
'release': v1_erratum['release'],
'rights': v1_erratum['rights'],
'severity': v1_erratum['severity'],
'solution': v1_erratum['solution'],
'status': v1_erratum['status'],
'summary': v1_erratum['summary'],
'title': v1_erratum['title'],
'type': v1_erratum['type'],
'updated': v1_erratum['updated'],
'version': v1_erratum['version'],
}
v2_coll.insert(new_erratum, safe=True)
# Throughout most of the upgrade scripts, they can be cancelled and
# resumed at any point and it will do the right thing. In this case,
# it's a nightmare to cross-reference the v1 erratum against the v2
# _id field. So adding the association here isn't 100% safe in the event
# the user ctrl+c's the upgrade (which they shouldn't do anyway) but
# it's close enough.
new_associations = []
for repo_id in v1_erratum['repoids']:
new_association = {
'_id': ObjectId(),
'repo_id': repo_id,
'unit_id': erratum_id,
'unit_type_id': 'erratum',
'owner_type': DEFAULT_OWNER_TYPE,
'owner_id': DEFAULT_OWNER_ID,
'created': DEFAULT_CREATED,
'updated': DEFAULT_UPDATED,
}
new_associations.append(new_association)
if new_associations:
v2_ass_coll.insert(new_associations, safe=True)
return True
def __init__(self):
self._id = ObjectId()
self.id = str(
self._id) # legacy behavior, would love to rid ourselves of this
def _repo_distributors(v1_database, v2_database, report):
v1_coll = v1_database.repos
v2_repo_coll = v2_database.repos
v2_dist_coll = v2_database.repo_distributors
# Only the yum distributor is added in this process. The export distributor
# will be added to any repositories that do not have it as part of the
# normal DB migration process.
# Idempotency: Only one distributor is added per repo, so check for that
v2_distributor_repo_ids = [x['repo_id'] for x in v2_dist_coll.find({}, {'repo_id' : 1})]
spec = {
'$and' : [
{'id' : {'$nin' : v2_distributor_repo_ids}},
{'content_types' : V1_YUM_REPO},
]
}
missing_v1_repos = v1_coll.find(spec)
new_distributors = []
for v1_repo in missing_v1_repos:
# Sanity check that the repository was added to v2. This should never
# happen, but we should account for it anyway.
v2_repo = v2_repo_coll.find_one({'id' : v1_repo['id']})
if v2_repo is None:
report.error('Repository [%s] not found in the v2 database; '
'distributor addition being canceled' % v1_repo['id'])
return False
new_distributor = {
'_id' : ObjectId(),
'id' : YUM_DISTRIBUTOR_ID,
'repo_id' : v1_repo['id'],
'distributor_type_id' : YUM_DISTRIBUTOR_TYPE_ID,
'auto_publish' : True,
'scratchpad' : None,
'last_publish' : v1_repo['last_sync'], # in v1 sync and publish are the same, so close enough
'scheduled_publishes' : [], # likely don't need to revisit, the sync/auto-publish will take care
}
config = {
'relative_url' : v1_repo['relative_path'],
'http' : False,
'https' : True,
}
# Load values from the static server.conf file
if not SKIP_SERVER_CONF:
parser = SafeConfigParser()
parser.read(V1_SERVER_CONF)
if parser.has_option('security', 'ssl_ca_certificate'):
# Read in the contents of the certificate and store in the
# configuration in the DB.
ca_filename = parser.get('security', 'ssl_ca_certificate')
try:
# Not bothering with a existence check, let any problems
# trigger the warning in the except
f = open(ca_filename, 'r')
ca_cert_contents = f.read()
f.close()
config['https_ca'] = ca_cert_contents
except:
report.warning('Could not read SSL CA certificate at [%s] for '
'repository [%s]' % (ca_filename, v1_repo['id']))
# Load the GPG keys from disk if present
repo_key_dir = os.path.join(GPG_KEY_ROOT, v1_repo['relative_path'])
if os.path.exists(repo_key_dir) and not SKIP_GPG_KEYS:
key_filenames = os.listdir(repo_key_dir)
if len(key_filenames) > 0:
filename = os.path.join(repo_key_dir, key_filenames[0])
try:
f = open(filename, 'r')
key_contents = f.read()
f.close()
config['gpgkey'] = key_contents
except:
report.warning('Could not read GPG key at [%s] for '
'repository [%s]' % (filename, v1_repo['id']))
new_distributor['config'] = config
new_distributors.append(new_distributor)
if new_distributors:
v2_dist_coll.insert(new_distributors, safe=True)
return True
def _repo_importers(v1_database, v2_database, report):
v1_coll = v1_database.repos
v2_repo_coll = v2_database.repos
v2_imp_coll = v2_database.repo_importers
# Idempotency: There is a single importer per repo, so we can simply check
# for an importer with the given repo ID
v2_importer_repo_ids = [x['repo_id'] for x in v2_imp_coll.find({}, {'repo_id' : 1})]
spec = {
'$and' : [
{'id' : {'$nin' : v2_importer_repo_ids}},
{'content_types' : V1_YUM_REPO},
]
}
missing_v1_repos = v1_coll.find(spec)
new_importers = []
for v1_repo in missing_v1_repos:
# Sanity check that the repository was added to v2. This should never
# happen, but we should account for it anyway.
v2_repo = v2_repo_coll.find_one({'id' : v1_repo['id']})
if v2_repo is None:
report.error('Repository [%s] not found in the v2 database; '
'importer addition being canceled' % v1_repo['id'])
return False
new_importer = {
'_id' : ObjectId(),
'id' : YUM_IMPORTER_ID,
'repo_id' : v1_repo['id'],
'importer_type_id' : YUM_IMPORTER_TYPE_ID,
'scratchpad' : None,
'last_sync' : v1_repo['last_sync'],
'scheduled_syncs' : [],
}
# The configuration intentionally omits the importer configuration
# values: num_threads, num_old_packages, remove_old, verify_checksum,
# verify_size, max_speed
# Being omitted will cause the yum importer to use the default, which
# is the desired behavior of the upgrade.
# All are set below. To keep consistent with a fresh install, the fields aren't
# defaulted to None but rather omitted entirely.
config = {}
if v1_repo['source']: # will be None for a feedless repo
config['feed_url'] = v1_repo['source']['url']
# Load the certificate content into the database. It needs to be written to the
# working directory as well, but that will be done in the filesystem scripts.
if v1_repo['feed_ca']:
if not os.path.exists(v1_repo['feed_ca']):
continue
f = open(v1_repo['feed_ca'], 'r')
cert = f.read()
f.close()
config['ssl_ca_cert'] = cert
if v1_repo['feed_cert']:
if not os.path.exists(v1_repo['feed_cert']):
continue
f = open(v1_repo['feed_cert'], 'r')
cert = f.read()
f.close()
config['ssl_client_cert'] = cert
# Load values from the static server.conf file
if not SKIP_SERVER_CONF:
parser = SafeConfigParser()
parser.read(V1_SERVER_CONF)
for o in ('proxy_url', 'proxy_port', 'proxy_user', 'proxy_pass'):
if parser.has_option('yum', o):
config[o] = parser.get('yum', o)
new_importer['config'] = config
new_importers.append(new_importer)
if new_importers:
v2_imp_coll.insert(new_importers, safe=True)
return True
def _sync_schedules(v1_database, v2_database, report):
v1_repo_collection = v1_database.repos
v2_repo_importer_collection = v2_database.repo_importers
v2_scheduled_call_collection = v2_database.scheduled_calls
# ugly hack to find out which repos have already been scheduled
# necessary because $size is not a meta-query and doesn't support $gt, etc
repos_without_schedules = v2_repo_importer_collection.find(
{'scheduled_syncs': {
'$size': 0
}}, fields=['repo_id'])
repo_ids_without_schedules = [
r['repo_id'] for r in repos_without_schedules
]
repos_with_schedules = v2_repo_importer_collection.find(
{'repo_id': {
'$nin': repo_ids_without_schedules
}}, fields=['repo_id'])
repo_ids_with_schedules = [r['repo_id'] for r in repos_with_schedules]
repos_to_schedule = v1_repo_collection.find(
{
'id': {
'$nin': repo_ids_with_schedules
},
'sync_schedule': {
'$ne': None
}
},
fields=['id', 'sync_schedule', 'sync_options', 'last_sync'])
for repo in repos_to_schedule:
if repo['id'] not in repo_ids_without_schedules:
report.error('Repository [%s] not found in the v2 database.'
'sync scheduling being canceled.' % repo['id'])
return False
args = [repo['id']]
kwargs = {'overrides': {}}
call_request = CallRequest(sync_with_auto_publish_itinerary,
args,
kwargs,
principal=SystemUser())
scheduled_call_document = {
'_id': ObjectId(),
'id': None,
'serialized_call_request': None,
'schedule': repo['sync_schedule'],
'failure_threshold': None,
'consecutive_failures': 0,
'first_run': None,
'last_run': None,
'next_run': None,
'remaining_runs': None,
'enabled': True
}
scheduled_call_document['id'] = str(scheduled_call_document['_id'])
schedule_tag = resource_tag(dispatch_constants.RESOURCE_SCHEDULE_TYPE,
scheduled_call_document['id'])
call_request.tags.append(schedule_tag)
scheduled_call_document[
'serialized_call_request'] = call_request.serialize()
if isinstance(repo['sync_options'], dict):
scheduled_call_document['failure_threshold'] = repo[
'sync_options'].get('failure_threshold', None)
interval, start, recurrences = dateutils.parse_iso8601_interval(
scheduled_call_document['schedule'])
scheduled_call_document['first_run'] = start or datetime.utcnow()
scheduled_call_document['remaining_runs'] = recurrences
scheduled_call_document['next_run'] = _calculate_next_run(
scheduled_call_document)
if repo['last_sync'] is not None:
scheduled_call_document[
'last_run'] = dateutils.to_naive_utc_datetime(
dateutils.parse_iso8601_datetime(repo['last_sync']))
v2_scheduled_call_collection.insert(scheduled_call_document, safe=True)
v2_repo_importer_collection.update(
{'repo_id': repo['id']},
{'$push': {
'scheduled_syncs': scheduled_call_document['id']
}},
safe=True)
return True
def test_expired_object_id(self):
expired_oid = self.reaper._create_expired_object_id(
timedelta(seconds=1))
self.assertTrue(isinstance(expired_oid, ObjectId))
now_oid = ObjectId()
self.assertTrue(now_oid > expired_oid)
