Summary

When linking a JIRA Issue to a remote link, such as a Confluence page, the JIRA issue is not reindexed

Steps to Reproduce

1. Re-index JIRA, ensuring that the Health Check shows the Lucene Index is healthy
2. Add debug logging on com.atlassian.jira.issue.index 
3. Insert a JIRA issues macro onto a new Confluence page (ensure that this is the first time that issue has been linked to that page)
4. Refresh the support tools screen - the index shows as unhealthy.
   • The database has received an update from Confluence, but has not been reflected in the index
5. Edit the issue that you linked, and it will be reindexed. Checking the support tools screen shows the lucene index is healthy again

Additional Diagnostic Information

Debug logging shows the following when an issue is edited and subsequently indexed:

2017-07-05 10:55:53,554 http-nio-9733-exec-25 DEBUG admin 655x693x1 bmvmt3 0:0:0:0:0:0:0:1 /secure/AjaxIssueAction.jspa [c.a.j.issue.index.DefaultIssueIndexer] Performing an index operation: All issues provided by iterable will be specifically reloaded from the database before indexing
2017-07-05 10:55:53,557 http-nio-9733-exec-25 DEBUG admin 655x693x1 bmvmt3 0:0:0:0:0:0:0:1 /secure/AjaxIssueAction.jspa [c.a.j.issue.index.DefaultIssueDocumentFactory] Indexing issue: TEST-11
2017-07-05 10:55:53,575 http-nio-9733-exec-25 DEBUG admin 655x693x1 bmvmt3 0:0:0:0:0:0:0:1 /secure/AjaxIssueAction.jspa [c.a.j.issue.index.DefaultIndexManager] Reindexed 1 issues in 21ms.

This debug logging does not appear for issues that have only been touched when updated via a remotelink. This behaviour can also be verified, by using Adaptavist ScriptRunner, which queries via the index:

ScriptRunner can be used to validate the problem. Use the following JQL:

issueFunction in linkedIssuesOfRemote("query", "pageId=xxx")

In this JQL, xxx is a page ID in Confluence. It will show all the issues that have a remote link back to that page ID - but if the issue has not been updated in the index, it cannot be found via this function (or other index based functions).

Issue Linking from Confluence can lead to issues where the database is up to date, but the Lucene index is not up to date, which makes searching and reporting on issues in JIRA unreliable. This will also affect all JQL queries based on the index - even viewing issues in issue navigator, ordered by updated date shows the issue hasn't been touched recently, when it actually has.

This also occurs when you link an issue to a Confluence page from inside JIRA - the index is not updated in the same fashion. 

Expected Results

The issue is reindexed, as if it has been updated manually.

Actual Results

The issue is not reindexed, and the Lucene Health Check fails 

Workaround

Touch the issue to perform a reindex, or reindex JIRA.