When syncing large numbers of users or groups from a remote directory, Crowd loads all updated users in memory when updating the cache, which can lead to {{OutOfMemoryError}}s.

All CacheRefresher implementations seem to follow the same pattern:

• retrieve all (updated) users from the remote directory
• for each user that needs to be added - create a new Crowd user in memory
• for each user that needs to be updated - create a new Crowd user in memory
• bulk create / update / delete the Crowd users in database.

A similar pattern is followed for the groups. If the remote directory has many users or groups, Crowd sync can use a lot of memory. We've seen 175 Mb in use in a Stash support case where an OutOfMemoryError was triggered.

The logic in the various CacheRefreshers should be updated to fetch a limited number of users/groups at a time (say 5000), process those and then move on to the next batch.