High load on systems with many users

Under certain circumstances systems with many users suffer from high load.

Scenario to reproduce

• Jira with many users (10K+)
• lot's of changes to some of the profile elements for all users (e.g. after first synchronization to LDAP)

Observations:

• the server is constantly under high load
• lots of entries in the service overview for the service which read like
  "Successfully reindexed 0 issues for 21344 changed users and 3 profile elements in 31636ms"
  • the important detail are the "0 issues"

Cause:

• there is a log of changed users which does not get flushed if there where no issues to reindex
• this causes the users to be processed again and again, resulting in high load