Unable to autocomplete some usernames in Jira Data Center

Summary

When typing usernames in Jira fields like Assignee, Reporter, or @mentioning users in comments, some usernames will show up in autocomplete, and others will constantly fail.

In the browser's console, you may see errors like:

"Exception: Uncaught TypeError: Cannot read properties of undefined (reading 'issues')"

Environment

All versions of Jira 8.x.

Diagnosis

You may observe this behavior for what seems to be all usernames or just some, while others work.

1) Log evidence

While reproducing the issue, watch Jira's atlassian-jira-log for a similar message:

The key is the "has no unique key mapping":

2) Database evidence

For each username that's returned in the grep above, check if they have the same lower_user_name in both tables below or if in one of them, there's a trailing #1 or other number:

Cause

If both the log and database evidences above are confirmed, it means Jira has lost consistency of this user's record. The lower_user_name of both tables app_user and cwd_user should always match.

Among possible root causes are:

• The user was renamed at some point and the change synced into Jira. Even if the change was reverted in the underlying User Directory, Jira wasn't able to recover on it's own

• Some duplicate username came in through the User Directory sync. Even if the duplicate was soon fixed, Jira wasn't able to figure out on it's own on further syncs.

• There were changes on key attributes to Jira in the User Directory, like a user's objectGUID or sAMAccountName and Jira wasn't able to resolve the conflict.

In all these situations, we'll see #1 or other numbers appended to usernames in Jira's database.

Solution

1) Full directory sync

The first advised approach is to force a full directory sync. You may trigger it by editing the LDAP directory, saving it without any changes and press the resync button.

Jira will understand that the directory has been recently edited and will force a full sync instead of trying an incremental one.

If the issue persists, try the next solution.

2) Manual intervention

We may manually fix the DB records with the following steps — though it requires some downtime. We also advise following the steps in a non-production instance first to validate it to your scenario and get used to the commands:

1. Stop Jira (all nodes).

2. Back up Jira's database to be used in a rollback/backout if needed.

3. Update the mismatched record(s). The example's for username "some-jira-user#1":

   1 update app_user set lower_user_name = 'some-jira-user' where lower_user_name = 'some-jira-user#1';

4. Commit the DB changes and start Jira.

Confirm if usernames show up when autocompleting and if there is no other similar evidence in the logs.