lastComment JQL function is not scalable for enterprise

Hi Team,

I'm raising this ticket on behalf of XXXXX. 

Here is XXXXX instance statistic, four node datacenter, @20GB xmx.

___ Database Statistics ____________________

     Issues                                        : 3361594
     Projects                                      : 1385
     Custom Fields                                 : 567
     Workflows                                     : 668
     Users                                         : 39783
     Groups                                        : 11505
     Attachments                                   : 1933253
     Comments                                      : 12177867

XXXXX reported multi-node crash due to OOME. from reviewing the heap dump generated upon OOME, we found lastComment function as the culprit. This function appears to load all Lucene comment documents into memory (JE: this is not the case in fact although clearly there is a problem) to perform the calculation. This array of comment is equivalent to 1.2-1.3 GB of heap memory.

So imagine there are 10 filters running this function concurrently which a 12-13 GB memory commitment.

Additionally, this function appears doesn't respect the Project scope given in the filter- it loads all comment despite the filter query.

Question.

• is this a known issue, anything you can suggest to XXXX to improve the performance of lastComment?
• is there any similar JQL function from ScriptRunner that can potentially cause the same problem? ie. lastWatcher etc

Best Regards,

Vicky Kharisma
Atlassian Premier Support