def test_not_found(self, mocked_load_search_apps):
"""
Test that get_search_app raises KeyError if it can't find the right
search app for the name passed in.
"""
mocked_load_search_apps.return_value = {
'app1': mock.Mock(),
}
with pytest.raises(KeyError):
get_search_app(mock.Mock())
def test_company_dbmodels_to_documents(self, opensearch):
"""Tests conversion of db models to OpenSearch documents."""
companies = CompanyFactory.create_batch(2)
app = get_search_app('company')
companies_qs = app.queryset.all()
result = SearchCompany.db_objects_to_documents(companies_qs)
assert len(list(result)) == len(companies)
def sync_model(search_app_name):
"""
Task that syncs a single model to Elasticsearch.
acks_late is set to True so that the task restarts if interrupted.
priority is set to the lowest priority (for Redis, 0 is the highest priority).
"""
search_app = get_search_app(search_app_name)
sync_app(search_app)
def test_found(self, mocked_load_search_apps):
"""
Test that get_search_app returns the right search app for the name passed in.
"""
search_app = mock.Mock()
mocked_load_search_apps.return_value = {
'app1': mock.Mock(),
'app2': search_app,
}
assert get_search_app('app2') == search_app
def sync_object_task(search_app_name, pk):
"""
Syncs a single object to Elasticsearch.
If an error occurs, the task will be automatically retried with an exponential back-off.
The wait between attempts is approximately 2 ** attempt_num seconds (with some jitter
added).
This task is named sync_object_task to avoid a conflict with sync_object.
"""
from datahub.search.sync_object import sync_object
search_app = get_search_app(search_app_name)
sync_object(search_app, pk)
def test_company_dbmodel_to_dict(self, opensearch):
"""Tests conversion of db model to dict."""
company = CompanyFactory()
app = get_search_app('company')
company_qs = app.queryset.get(pk=company.pk)
result = SearchCompany.db_object_to_dict(company_qs)
keys = {
'_document_type',
'archived',
'archived_by',
'archived_on',
'archived_reason',
'business_type',
'company_number',
'created_on',
'description',
'employee_range',
'export_experience_category',
'export_to_countries',
'future_interest_countries',
'headquarter_type',
'id',
'modified_on',
'name',
'global_headquarters',
'reference_code',
'sector',
'latest_interaction_date',
'address',
'registered_address',
'one_list_group_global_account_manager',
'trading_names',
'turnover_range',
'uk_based',
'uk_region',
'uk_address_postcode',
'uk_registered_address_postcode',
'vat_number',
'duns_number',
'website',
'export_segment',
'export_sub_segment',
}
assert set(result.keys()) == keys
def complete_model_migration(self, search_app_name, new_mapping_hash):
"""
Completes a migration by performing a full resync, updating aliases and removing old indices.
"""
search_app = get_search_app(search_app_name)
if search_app.es_model.get_target_mapping_hash() != new_mapping_hash:
warning_message = f"""Unexpected target mapping hash. This indicates that the task was \
generated by either a newer or an older version of the app. This could happen during a blue-green \
deployment where a new app instance creates the task and it's picked up by an old Celery instance.
Rescheduling the {search_app_name} search app migration to attempt to resolve the conflict...
"""
logger.warning(warning_message)
raise self.retry()
with advisory_lock(f'leeloo-resync_after_migrate-{search_app_name}',
wait=False) as lock_held:
if not lock_held:
logger.warning(
f'Another complete_model_migration task is in progress for the {search_app_name} '
f'search app. Aborting...', )
return
resync_after_migrate(search_app)