[
https://issues.apache.org/jira/browse/OAK-2972?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14580088#comment-14580088
]
Chetan Mehrotra commented on OAK-2972:
--------------------------------------
The bug is in the activate logic
{code}
--&& (dataSource == null || (customBlobDataSource && blobDataSource == null))) {
++&& (dataSource == null || blobDataSource == null))) {
{code}
One possible fix would be to drop the check for {{customBlobDataSource}} as in
any case we want both to be initialized. If a {{customBlobDataSource}} is not
set then {{blobDataSource == dataSource}}. This would ensure that
initialization would be delayed untill all the required references are met.
Thinking more about it we should do away with {{customBlobDataSource}} config.
If a different DataSource has to be used then admin would change the target
filter for {{blobDataSource}} and activate would delay the init untill both are
defined
> DocumentNodeStore gets initialized multiple time with RDB persistence
> ---------------------------------------------------------------------
>
> Key: OAK-2972
> URL: https://issues.apache.org/jira/browse/OAK-2972
> Project: Jackrabbit Oak
> Issue Type: Bug
> Components: rdbmk
> Reporter: Chetan Mehrotra
> Labels: osgi
> Fix For: 1.3.1, 1.2.3, 1.0.15
>
>
> Depending on startup sequence it can happen that {{DocumentNodeStore}} gets
> initialized multiple times.
> So far with Mongo {{DocumentNodeStoreService}} was only dependent on one
> external reference of {{BlobStore}}. However with RDB there are two more
> external reference for {{DataSource}} one for nodes and other for blobs.
> SCR (Felix Service Component Runtime) would invoke both {{bindDataSource}}
> and {{bindBlobDataSource}} and currently there is no check to avoid re
> initialization.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
