[
https://issues.apache.org/jira/browse/OAK-805?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chetan Mehrotra updated OAK-805:
--------------------------------
Attachment: OAK-805.patch
Patch which reworks the implementation done for {{DataStoreBlobStore}} as part
of OAK-1157
Key points
* DataStoreBlobStore is rewritten so better to see it
[directly|https://github.com/chetanmeh/jackrabbit-oak/blob/OAK-1502/oak-core/src/main/java/org/apache/jackrabbit/oak/plugins/blob/datastore/DataStoreBlobStore.java]
rather than in diff
** Added support for inlining documents. In JR2 this was handled in layers
above DataStore. However for Oak this needs to be handled in BlobStore
** GC logic has to be reviewed
* Added new method in BlobStore {{InputStream getInputStream(String blobId)}}.
It enables efficient handling of InputStream for case where BlobStore is backed
by DataStore
* Replaced DocumentBlob with BlobStoreBlob as same Blob is also used in
SegmentStore (OAK-1333)
Queries
* Probably a better name for {{BlobStoreBlob}}
* GC handling has to be reviewed. Currently some methods in
org.apache.jackrabbit.oak.spi.blob.GarbageCollectableBlobStore are not related
to GC and can be removed (setBlockSize,writeBlob etc)
/cc: [~tmueller] , [~amitjain] Can you review the patch
See on gihub
[here|https://github.com/chetanmeh/jackrabbit-oak/commit/1cc25582583eeb310f151762aaa5234c292d8ec0]
> Support for existing Jackrabbit 2.x DataStores
> ----------------------------------------------
>
> Key: OAK-805
> URL: https://issues.apache.org/jira/browse/OAK-805
> Project: Jackrabbit Oak
> Issue Type: New Feature
> Components: core, mk, mongomk
> Reporter: Jukka Zitting
> Assignee: Chetan Mehrotra
> Fix For: 0.19
>
> Attachments: OAK-805.patch
>
>
> To speed up migrations from Jackrabbit 2.x to Oak and to allow multiple Oak
> repositories to share binaries like what's possible with the Jackrabbit 2.x
> DataStore mechanism, it would be good if our MKs, most notably MongoMK and
> SegmentMK, had a way to reference binaries stored in a DataStore instead of
> in the native blob store of the MK.
> The downside of such setups is of course that it splits up the garbage
> collection process, but the benefits of a speedier and less expensive
> migration will likely be worth it at least in some cases with huge amounts
> (TBs) of existing content.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
