[
https://issues.apache.org/jira/browse/OAK-9015?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17088626#comment-17088626
]
Nitin Gupta commented on OAK-9015:
----------------------------------
Merged in trunk : [https://svn.apache.org/viewvc?view=revision&revision=1876791]
After discussion on PR [https://github.com/oak-indexing/jackrabbit-oak/pull/133]
> ElasticserachIndexWriter#close method implementation doesn't do what the
> underlying contract expects | BulkProcessor resource is never closed and it
> always returns false
> -------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: OAK-9015
> URL: https://issues.apache.org/jira/browse/OAK-9015
> Project: Jackrabbit Oak
> Issue Type: Bug
> Reporter: Nitin Gupta
> Assignee: Nitin Gupta
> Priority: Major
>
> A new ElasticserachIndexWriter instance is created everytime the async
> indexer service runs and the close() method in the FullTextWriter interface
> has the following contract
>
> {code:java}
> /**
> * Closes the underlying resources.
> *
> * @param timestamp timestamp to be used for recording at status in
> NodeBuilder
> * @return true if index was updated or any write happened.
> */
> boolean close(long timestamp) throws IOException;
> {code}
> However the implementation in ElasticserachIndexWriter returns false always
> and doesn't closes the BulkProcessor instances created.
>
> The problem that needs to be handled here is that the bulk processor calls to
> elastic server are async in nature and the close method in
> ElasticserachIndexWriter could be called before the bulkProcessor completed
> it work. So we need to find a safe way to close the bulkProcessor and return
> true/false from ElasticserachIndexWriter#close() depending upon if the index
> was updated or not.
>
> Looking at [0], both the bulkProcessor.close and awaitClose methods close the
> bulkPocessor after flushing all the documents that were loaded to the
> bulkProcessor till that moment. So we just need to make sure that, all the
> needed docs should be loaded in the bulkProcessor when
> ElasticserachIndexWriter#close() is called.
>
> [0]:
> [https://www.elastic.co/guide/en/elasticsearch/client/java-api/current/java-docs-bulk-processor.html]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
