[
https://issues.apache.org/jira/browse/OAK-2598?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14354588#comment-14354588
]
Alex Parvulescu commented on OAK-2598:
--------------------------------------
bq. To make it automatic you would need to find out such nodes and modify the
property before running the migration and then again query for such nodes after
running the migration.
JR2 only had a global index, so if we're considering adding the index
definitions in a pre-upgrade step, the upgrade component already knows what
definitions it added without the need to look for them after, here it can also
strip out the async markers. I think this info can also be persisted in a
hidden node to be picked up later (:upgrade ?).
In a post-upgrade step once everything is done, it can add the markers back in,
create a checkpoint and the good reference so everything looks good from the
async indexer's point of view.
> Provide option to run async index as sync in repository upgrade
> ---------------------------------------------------------------
>
> Key: OAK-2598
> URL: https://issues.apache.org/jira/browse/OAK-2598
> Project: Jackrabbit Oak
> Issue Type: Improvement
> Components: upgrade
> Reporter: Chetan Mehrotra
> Fix For: 1.1.8, 1.0.13
>
>
> Currently when performing repository upgrade from JR2 to Oak the migration
> logic only runs the synchronous index editors. Async indexes like Lucene/Solr
> are run post migration after the system start.
> Given that migration is a single threaded operation it would at times be
> helpful to allow executing such async indexes in sync mode during migration
> phase. This would avoid rescanning of complete repository again for such
> async indexes
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
