Issue Summary

Misbehaviour of the single issue reindexing API call: triggers issue and all related entities reindex instead of reindexing just the issue.

Steps to Reproduce

Jira is exposing a single issue re-index API:

// com.atlassian.jira.issue.index.IssueIndexManager
/**
 * Reindex an issue (eg. after field updates).
 */
void reIndex(Issue issue) throws IndexException;


// com.atlassian.jira.issue.index.IssueIndexingService

/**
 * Reindex an issue (eg. after field updates).
 */
void reIndex(final Issue issue) throws IndexException;

According to the javadoc this method should trigger an issue re-index but currently it is also re-indexing all related entities (comments, worklogs):

• bump issue version
• bump this issue comments versions
• bump this issue worklogs versions
• conditionally update local issue, comments, worklogs index (note that for comments and worklogs this will trigger a conditional index update for every comment/worklog separately)
• (Jira DC) send the documents (issue, comments, worklogs) via DBR to all other nodes and conditionally update other nodes indexes
• (Jira DC) add indexing operation (issue, comments, worklogs) to the index replication log
• (Jira DC) other nodes should process the index replication log and conditionally re-index this issue, comment, version 

If the issue contains a large amount of related entities this may have serious indexing implication (tens of thousands of index update operations across the cluster instead of a single index update across the cluster). 

Expected Results

According to the javadoc this method should trigger an issue re-index only:

• bump issue version
• conditionally update local issue index
• (Jira DC) send the document (issue only) via DBR (with issue only)to all other nodes and conditionally update other nodes indexes
• (Jira DC) add indexing operation (issue only) to the index replication log
• (Jira DC) other nodes should process the index replication log and conditionally re-index (issue only)

Actual Results

Timeouts on the comments and indexing queues may occur. 

Workaround

• Change on the API caller to specifically only re-index the issue:

  // com.atlassian.jira.issue.index.IssueIndexingService
  
  /** 
   * Reindex an issue (eg. after field updates). 
   */ 
  void reIndex(final Issue issue, final IssueIndexingParams issueIndexingParams) throws IndexException;
  
  // where issueIndexingParams = INDEX_ISSUE_ONLY | INDEX_ISSUE_WITH_HISTORY 

• Archive issue with "large" number of related entities
• In version 8.16.2, 8.17.1, >=8.18.0 one can set a flag to disable this behaviour:

-Dcom.atlassian.jira.issue.reindex.legacy.mode=false

With this flag the single issue re-index API:

// com.atlassian.jira.issue.index.IssueIndexManager 
/** 
 * Reindex an issue (eg. after field updates). 
 */ 
void reIndex(Issue issue) throws IndexException; 

// com.atlassian.jira.issue.index.IssueIndexingService  
/** 
 * Reindex an issue (eg. after field updates). 
 */ 
void reIndex(final Issue issue) throws IndexException;

is re-indexing the issue only.

This should improve the stability of the whole Jira DC cluster.