Versioning Operations
Note
Operations referencing the current version proceed normally if version={etag}
is used where versioning is disabled, but any other ETag results in an HTTP 404 - Not Found. (v9.2)
Adding Versions
Any write operation (POST, PUT, COPY, APPEND) in a versioning-enabled context creates a new version, and it changes the existing version to be the prior one. (Prior versions are not modified. All write operations create new versions.)
POST stores a brand new object as the current version. (PUT can be used if you have
scsp.allowPutCreate
enabled or add theputcreate
query argument.)PUT adds a new version of the object, replacing the current version (which remains accessible by ETag); use PUT to update an existing alias. You cannot use PUT with the
version
query argument.COPY ?version={etag} creates a new version duplicating the content of a versioned object with the new headers specified.
APPEND ?version={etag} creates a new version duplicating the headers of a versioned object with the new content specified (use 0-length to duplicate the existing content).
Retrieving Versions
Important
Consider each version being added as being an entire object for the purpose of resource management: it is not a delta from the previous version. Object counts and space utilization reflect 12 objects if you have 12 versions of an object stored (the current and 11 prior).
There are different ways to retrieve current objects versus specific object versions. Add the version query argument to retrieve a specific version:
GET retrieves the current version of an object.
GET ?version={etag} retrieves a specific version of an object, identified by its ETag header value.
HEAD retrieves the metadata of the current version of an object (and not the content).
HEAD ?version={etag} retrieves the metadata of a specific version of an object, identified by its ETag.
Specify the object's ETag to retrieve the content or metadata of a specific version, which retrieves the specified version of the object. This can be the current version or any prior version.
An HTTP 404 Not Found response including the ETag for the delete marker (not the prior version) is received if a GET or HEAD is issued against the delete marker:
Listing Versions
With Elasticsearch indexing your cluster, use it to access listings of your object versions. (Elasticsearch is not required to use versioning.)
Versioning headers – Swarm returns these additional headers with historical (non-current) versioned objects:
uuidNextVersion – Points to the primary UUID of the object that is the newer version. This field transfers during replication within the cluster but not to remote clusters (each cluster resolves its own chains).
tmNextVersion – Holds the birth date of the newer (uuidNextVersion) object (they update as a pair). This field transfers during replication within the cluster but not to remote clusters.
tmDeleted – Holds the death date of this particular version. This field does transfer to remote clusters so the needed feed processing and clean up occurs.
Differences from S3
Unlike Amazon S3 (which retrieves only a maximum of 1000 objects), Swarm has no such limits, nor does it break batches on version boundaries.
With rapid updates of versioned objects, there is a delay in listing consistency, which can cause 409 Conflict and 503 Service Unavailable responses. Using
replicate=immediate
does not prevent 409 responses; a 1-second delay always applies to versioned buckets.
Add the versions query argument to the GET Bucket request to list all versions of all objects in a bucket.
Use prefix query argument along with versions query argument to the GET Bucket request to list all versions of one object, which limits the set of versions returned to those related to the object.
GET /mybucket?domain=sample&format=json&versions HTTP/1.1
GET /mybucket?domain=sample&format=json&versions&prefix=objectName HTTP/1.1
Deleting Versions
The behavior is the same when versioning is enabled or suspended:
Deleting without specifying a version: Swarm operates on the current version and creates a delete marker.
Deleting with a version specified: Swarm permanently pushes the version out of existence, even if it is the current version.
A DELETE does not permanently delete an object when versioning is enabled. Swarm inserts a delete marker instead, and the marker becomes the current version of the object with a new ID.
The response to a DELETE operation includes the ETag of the version being deleted, not that of the delete marker:
Use DELETE Object and specify the ETag to permanently delete versioned objects:
DELETE /mybucket/photo.gif?version=c347edc55b3c4fadb76d93022a29b07a HTTP/1.1
Handling Delete Markers
A delete marker is a placeholder (marker) for a versioned object named in a DELETE request. Because the object was in a versioning-enabled bucket, the object was not deleted. The delete marker causes Swarm to operate as if it had been deleted. A delete marker has a name and version ETag like any other object, but it differs from other objects in several ways:
It has no data.
Its only operation is DELETE, which only the owner can request.
It has nominal size.
Swarm can create a delete marker, and it does so when a DELETE Object request is sent on a versioning-enabled/suspended object . The object named in the DELETE request is not actually deleted. Instead, the delete marker becomes the current version of the object. (The object's name becomes the name of the delete marker.)
The version ETag in a DELETE Object request must be included to permanently delete a delete marker. Owners can permanently delete these markers while the context is versioning-enabled/suspended. (The health processor cleans up the markers along with the prior versions if the context returns to versioning-disabled.)
Lifepoints for Versions
Object versions can be deleted on demand, but lifepoints can be defined for objects with a well-defined lifecycle to have Swarm permanently remove prior object versions as they expire. The Health Processor cleans up (erases) versions with expired lifepoints for both versioning-enabled and versioning-suspended states, deleting particular object versions. The Heath Processor repairs the gap in the version sequence caused by the deletion.
The Health Processor creates a delete marker that becomes the current version when the lifepoint expires on the current version of the object: it does not resurrect a previous version.
The Health Processor acts on versioned objects in these ways:
Evaluates lifepoints of current versions, performing EC-related conversions or lifepoint deletions.
Keeps the desired number of replicas of versioned objects.
Maintains the current version of an object and the chain of version linkages.
Deletes prior versions of objects no longer accessible.
Renaming Versioned Objects
Renaming a versioned object is performed by ending the old one and starting a new one. There are two requests and two objects:
a DELETE on the old name, which terminates the version chain with a delete marker
a WRITE of a new object, with the new name
Note: the name change can be reversed by deleting the new object, reading the last good version, and appending the version.
Restoring Versions
A key benefit of versioning is the ability to restore previous versions of an object. There are several ways to restore content by referencing specific versions, depending on your situation:
Problem | Method | Reason |
---|---|---|
The last update was erroneous | Destroys: Permanently delete the current version by ETag: Swarm automatically promotes the previous version to be the current version of the object when deleting the current object by the version ETag. | |
An older version must be restored | Preserves: Copy a previous version of the object into the same context to restore any one of the prior versions. The copied object becomes the current version of the object, and all object versions are preserved. APPEND adds data onto the end of an existing named object or aliased object while maintaining all populated metadata and object name or alias UUID. Use a 0-length APPEND to duplicate the prior version exactly without change to content or metadata. Any earlier version can be made the current version by copying a specific version of the object into the same bucket because all object versions are preserved. Swarm supplies a new ID and it becomes the current version of the object. A subsequent GET retrieves the copy. | |
New metadata needs to be added | Replaces: Use COPY to duplicate the prior version with updates to the metadata. COPY updates the metadata on an existing object by copying the content verbatim while replacing the metadata. This technique facilitates extending the custom object metadata over time, to improve the scope and usefulness of your Elasticsearch queries. | |
An older version must be updated | PUT | Invalid: Prior versions are historical and cannot be altered (only deleted), so PUT cannot be used with them. Use COPY or APPEND on the older version, including the changes needed. Use DELETE to remove the specific version permanently if the existence of the incorrect version is a concern or a liability. |
© DataCore Software Corporation. · https://www.datacore.com · All rights reserved.