return self._get_manifest_text(stream_name, strip, normalize)
@synchronized
- def _get_manifest_text(self, stream_name, strip, normalize):
+ def _get_manifest_text(self, stream_name, strip, normalize, only_committed=False):
"""Get the manifest text for this collection, sub collections and files.
:stream_name:
is not modified, return the original manifest text even if it is not
in normalized form.
+ :only_committed:
+ If True, only include blocks that were already committed to Keep.
+
"""
if not self.committed() or self._manifest_text is None or normalize:
for segment in arvfile.segments():
loc = segment.locator
if arvfile.parent._my_block_manager().is_bufferblock(loc):
+ if only_committed:
+ continue
loc = arvfile.parent._my_block_manager().get_bufferblock(loc).locator()
if strip:
loc = KeepLocator(loc).stripped()
num_retries=None,
parent=None,
apiconfig=None,
- block_manager=None):
+ block_manager=None,
+ replication_desired=None):
"""Collection constructor.
:manifest_locator_or_text:
a manifest, raw manifest text, or None (to create an empty collection).
:parent:
the parent Collection, may be None.
+
:apiconfig:
A dict containing keys for ARVADOS_API_HOST and ARVADOS_API_TOKEN.
Prefer this over supplying your own api_client and keep_client (except in testing).
Will use default config settings if not specified.
+
:api_client:
The API client object to use for requests. If not specified, create one using `apiconfig`.
+
:keep_client:
the Keep client to use for requests. If not specified, create one using `apiconfig`.
+
:num_retries:
the number of retries for API and Keep requests.
+
:block_manager:
the block manager to use. If not specified, create one.
+ :replication_desired:
+ How many copies should Arvados maintain. If None, API server default
+ configuration applies. If not None, this value will also be used
+ for determining the number of block copies being written.
+
"""
super(Collection, self).__init__(parent)
self._api_client = api_client
self._keep_client = keep_client
self._block_manager = block_manager
+ self.replication_desired = replication_desired
if apiconfig:
self._config = apiconfig
def _my_api(self):
if self._api_client is None:
self._api_client = ThreadSafeApiCache(self._config)
- self._keep_client = self._api_client.keep
+ if self._keep_client is None:
+ self._keep_client = self._api_client.keep
return self._api_client
@synchronized
@synchronized
def _my_block_manager(self):
if self._block_manager is None:
- self._block_manager = _BlockManager(self._my_keep())
+ copies = (self.replication_desired or
+ self._my_api()._rootDesc.get('defaultCollectionReplication',
+ 2))
+ self._block_manager = _BlockManager(self._my_keep(), copies=copies)
return self._block_manager
def _remember_api_response(self, response):
uuid=self._manifest_locator).execute(
num_retries=self.num_retries))
self._manifest_text = self._api_response['manifest_text']
+ # If not overriden via kwargs, we should try to load the
+ # replication_desired from the API server
+ if self.replication_desired is None:
+ self.replication_desired = self._api_response.get('replication_desired', None)
return None
except Exception as e:
return e
ensure_unique_name = True
body = {"manifest_text": text,
- "name": name}
+ "name": name,
+ "replication_desired": self.replication_desired}
if owner_uuid:
body["owner_uuid"] = owner_uuid