[
https://issues.apache.org/jira/browse/OAK-9004?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17082120#comment-17082120
]
Amrit Verma commented on OAK-9004:
----------------------------------
PR link - [https://github.com/oak-indexing/jackrabbit-oak/pull/117/files]
> Multitenant index name strategy with reindex support
> ----------------------------------------------------
>
> Key: OAK-9004
> URL: https://issues.apache.org/jira/browse/OAK-9004
> Project: Jackrabbit Oak
> Issue Type: Task
> Components: indexing
> Reporter: Amrit Verma
> Priority: Major
>
> Index names in elastic follow the pattern *<name>-<reindex-count>*
> This has to be changed in order to support multi-tenancy.
> We need to pass a parameter with the customer ID so we can create indexes
> like *<customer>.<name>-<reindex-count>*. ES indexes cannot be longer than
> 255 bytes and must comply with the following criteria
> [https://www.elastic.co/guide/en/elasticsearch/reference/current/indices-create-index.html#indices-create-api-path-params]
>
> We need also to decide what to do on re-index. The simpler strategy would be
> to:
> * create the new index
> * move writes and reads to the new index
> * delete old index
> A more sophisticated strategy could:
> * create the new index
> * move writes to the new index
> * reads will continue to use the old index until the new one catches up
> * when the new one is in sync, move reads to the new index & delete the old
> one
> Both strategies can be implemented using Aliases in Elasticsearch to avoid
> race conditions. To implement the second solution we need something that
> tells us when the new index has caught up with the initial load.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
