-
Bug
-
Resolution: Unresolved
-
Low
-
None
-
5.9.1, 5.8.10, 5.9.5, 5.9.7, 5.10.8
-
23
-
Severity 3 - Minor
-
2
-
NOTE: This bug report is for Confluence Server. Using Confluence Cloud? See the corresponding bug report.
Summary
In the current System Properties documentation there is a setting officeconnector.textextract.word.docxmaxsize, and this is ignored by Confluence when set in setenv.sh.
Environment
- Confluence 5.8.x or Confluence 5.9.x
Steps to Reproduce
- Add to setenv.sh
setenv.sh
CATALINA_OPTS="-Dofficeconnector.textextract.word.docxmaxsize=1000 ${CATALINA_OPTS}" - Insert the attached file lorum.docx to a page.
- Check the logs, the error
atlassian-confluence.log
com.atlassian.bonnie.search.extractor.ExtractorException: Error reading content of Word XML document: Document too big for text extraction, bailing outdoes not appear.
Expected Results
The error
com.atlassian.bonnie.search.extractor.ExtractorException: Error reading content of Word XML document: Document too big for text extraction, bailing out
should appear as the file size is now 1K.
Actual Results
The file is indexed correctly.
Notes
For the attached file you will need to increase your Java Heap Space in setenv.sh to something like -Xmx8192m.
The value officeconnector.textextract.word.docxmaxsize is referenced in com.atlassian.confluence.extra.officeconnector.index.word#WordXMLTextExtractor.java as
private static final long MAX_XML_SIZE = Long.getLong("officeconnector.textextract.word.docxmaxsize", 1024 * 1024 * 16); // Maximum, 16Mb of XML text, more than enough for most files ... if (finalSize > MAX_XML_SIZE) { throw new ExtractorException("Document too big for text extraction, bailing out"); }
Workaround
There is no workaround.
- is related to
-
-
- Closed
-
- relates to
-
-
- Closed
-
- Testing discovered
-
CONFSERVER-40432 Filter Out All Media Files from Microsoft Word Documents to Improve Indexing in Confluence
- Closed
- mentioned in
-
![[Confluence] Page](/images/icons/generic_link_16.png "[Confluence] Page")
Page
No Confluence page found with the given URL.
-
Page Loading...
-
-
-
-
-
-
-
-
-
-
-
-
-
![[Confluence] Page](https://bulldogwiki.internal.atlassian.com/wiki/images/icons/favicon.png "[Confluence] Page"){kind=icon}
[CONFSERVER-40176] Confluence ignores the system property officeconnector.textextract.word.docxmaxsize
Thanks for doing that testing, Guilherme! The memory usage from this bug is a little nerve-wracking, so I appreciate you getting to this so quickly.
Thanks James, I will do a bit more testing to see if I can narrow this down any further.
The documents are 8.53 MB, 17.44 MB, and 6.08 MB, with no video content, so it just doesn't seem like they should be running into this.
Hi adaly,
It's 16MB of uncompressed text. The code strips out a few multimedia formats (see CONF-40432), so if you have embedded video it may then jump to the higher values, but it really depends on the document size and contents.
If it's an issue, please log a ticket with support and we can investigate further.
James.
What is the default value for this system property?
I ask because with Confluence 5.9.7, I have three docx files that throw this error.
When this occurs, system memory usage immediately jumps 2 gig (yes, gig) and does not return that memory until I restart the service.
Same issue here on Confluence 5.9.10