Summary

All HTTP requests in Cloud have a 60 second timeout. When a PDF export runs longer than that, the task keeps running on the background but the .pdf generated is not returned to the customer. If another PDF export is triggered, Confluence tries to generate a new PDF, even though it already has one from the last export and again, it wont be downloadable.

There is also a 30 second retry done automatically by the smart-status timeout page from the client browser that will trigger new PDF exports. Confluence will deny a new PDF export to run with the message:

java.lang.IllegalStateException: Too many concurrent exports. 1 is max allowed.

but as soon as the export finishes it will start a new PDF export.

PDF exports that take longer than 1 minute are high CPU tasks. This loop of generating new pdf files can render Cloud sites unusable on certain situations.

Environment

• Confluence Cloud

Steps to Reproduce

1. Create a Confluence Cloud page with 100 external images from another Cloud site (this will fail to retrieve images for the pdf at the rate of 1 image each second.
2. Export page to PDF

Expected Results

pdf file is downloadable at the end of the process. If it is not, then a new PDF export downloads the previous generated file if the page wasn't modified.

Actual Results

Confluence generates a pdf indefinitely until the browser is stopped.
New PDF exports are rejected but as soon as it completes, it starts a new one when it is automatically requested by the smart-status 30 second refresh.

confluence/home/temp directory contains multiple .pdf files of the same page

Suggestion

Confluence could detect that there is a pdf export already in the temp directory for the same unmodified page and transfer that to the browser or provide the option for the customer to decide; in case there is dynamic content on the page and a new pdf export is explicitly requested.

Workaround

• Use the Space Tools PDF Export Option: as a Space Admin, Space Tools > Content Tools > Export