Confluence performance issues due to Scroll Word Export page
Platform Notice: Data Center Only - This article only applies to Atlassian products on the Data Center platform.
Note that this KB was created for the Data Center version of the product. Data Center KBs for non-Data-Center-specific features may also work for Server versions of the product, however they have not been tested. Support for Server* products ended on February 15th 2024. If you are running a Server product, you can visit the Atlassian Server end of support announcement to review your migration options.
*Except Fisheye and Crucible
Summary
Confluence experiencing performance issues due to Scroll Word Page Exports.
Environment
Confluence Data Center version 7.13.2
Diagnosis
From the application logs, we can clearly observe issues around Scroll Word Page Export which bumped into errors
1
2
3
4
5
6
7
XXXX-XX-XX XX:XX:XX,922 WARN [scroll-office-export:thread-2] [scroll.words.exporter.WordsExporter] exportPage Error when rendering replacement '$scroll.metadata.(rev)'
– pageId: XXXXXXXX | url: /plugins/servlet/scroll-office/api/exports | traceId: af68e2812e72f4da | userName: XXXXXXX | jobId: 9285ab7a-8bcf-437e-876a-18f027b686c5
com.k15t.scroll.exporter.api.replacement.ReplacementException: No metadata value for key 'rev'.
at com.k15t.scroll.exporter.api.replacement.MetadataPlaceholder.invoke(MetadataPlaceholder.java:125)
at com.k15t.scroll.words.exporter.WordsExporter.exportPage(WordsExporter.java:414)
at com.k15t.scroll.words.exporter.WordsExporter.export(WordsExporter.java:375)
...
Also, from the GC logs, we can see big memory pressure (Full GC peaks are uneven) under Confluence around the same time and under extreme memory pressure cases, this may also cause to observe Fake OOM error as documented under - Atlassian Support Jira Software 8.22 JIRA Knowledge Base JIRA Application Administration Knowledge Base JIRA Running out of memory due to GC / Allocation race condition.

Cause
The Scroll Page Export request crashed for some reason but then the memory related to it got leaked and we can observe a large chunk of heap occupied by Scroll Page Export requests under heap dump. Due to this, as more and more requests end up with this behavior, we can see a lot of memory pressure and Confluence performance will degrade with time.

Solution
Please reach out to the Scroll Word Export Support Team to check this and in case this is frequently causing performance issues, see if we can disable this plugin temporarily. You can also refer to some control measures as shared by the Scroll Team i.e and check if this helps.
Was this helpful?