We identified 3 use cases:
1 - Delete a document
2 - Edit a document: edition of a blob field, need to clean the blob that was referenced.
3 - Moving a blob from one blob store to another.
A dry run should be necessary to have a full GC logs before launching a hard delete.
A repo scan will be necessary to make sure that a blob isn't referenced elsewhere before being deleted. Following NXP-29516, we need to use the blob key ecm:blobKeys.
What happen for the documents created before this fix? NXP-30070 should resolve the issue. Note that this ticket depends on NXP-26909.
Epic Milestones:
1. Migration service: NXP-26909
2. Real-time orphan binaries cleanup. Make sure that when we delete a doc or edit a blob field it isn't referenced elsewhere.
3. Full GC, using works from milestone 2 and implementing scrollers. In blob store: scroll over them and reapply implementation done in milestone 2.
- is related to
-
NXP-32044 Make BlobProvider specification the default binary manager
- Resolved
-
NXP-31991 Provide a GoogleBlobProvider extending BlobProvider
- Resolved
-
NXDOC-2561 Document Full and Immediate GC
- Resolved
-
NXP-28565 Make orphan binaries GC scalable
- Resolved
-
NXP-31990 Provide a AzureBlobProvider extending BlobProvider
- Resolved
-
NXP-32046 Provide a GridFSBlobProvider extending BlobProvider
- Resolved
-
NXDOC-2591 Document how to ensure traceability for binaries
- Resolved