Current behavior

The catch up is performed in a single thread.

Expected behavior

The catch up is performed in multiple threads which is faster than in single thread.

Notes

Since 9.1 the index startup flow consists of 3 steps:

1. Re-index missing data if the local issue index is less than 10% behind the database. Since 9.3 the % is configurable.
2. Load a recent index snapshot from shared-home directory if one is available.
3. Otherwise trigger a full re-index.

(From https://confluence.atlassian.com/jirakb/index-management-on-jira-start-up-1141500654.html)

Step 1 and step 2 (after indexes from snapshot are unarchived and restored) both perform a catch up that re-indexes issues that were modified but not reflected in index state

Currently the catch up happens in a single thread unlike during full foreground reindex.

The multi-threaded catch up will cut the time of node startup, especially for deployments where Lucene documents creation takes a long time.