Confluence Queue Contents not flushing on own

[Patricia Wiebe]

Hi,

Anyone else seen this? We are running Confluence 4.2.2. Today we noticed that things were not updating on our pages and blogs. We checked the Queue Contents and it was not flushing on it's own. We have a scheduled job to flush every minute, and the task says it's doing it (as long as we are talking about the "Flush Index Queue), but the queue is not being flushed. Oh, and the mail queue also isn't flushing. Any thoughts - anyone seen this happen?? It just seemed to stop on it's own, we have not had any problems before this.

Thanks!

[Patricia Wiebe · Accepted solution]

I think we solved this. I restarted confluence and then did a search re-index. At 98% of the re-index, an exception was thrown regarding the "too many files open" problem.

20:54:59,954 ERROR [STDERR] org.apache.lucene.index.MergePolicy$MergeException: java.io.FileNotFoundException: /export/atlassian/confluence/4.2.2/core/data/temp/lucene5020753486088339503index/_8h.fnm (Too many open files)

We were familar with this problem, but thought it had been addressed with our newer version of confluence (we went from 3.3.1 to 4.2.2). Also for reference, we run on JBoss 4.2.3 and are running Confluence 4.2.2 with Oracle 11g.

https://confluence.atlassian.com/x/ygCmC

So, I added to our start script: ulimit -n 10000

After that restart, I did the re-index and it completed successfully with no errors. Since that restart, the queues have been flushing normally on the 1 minute schedule.

I'm not completely sure why the open file limit would affect the queue flushing, and it could have been just the restarts that did the trick, but we are running normally again.

[Leena Bakshi]

When we restarted the confluence , it fixed the issue for two days and today on the third day it is back again.. and I did rebuild the index from scratch also after restarting the service. any idea about the resolution.. 

[Sow Behl]

We just restarted Confluence and it seemed to fix the issue.

[Matthew J. Horn]

I saw this with comments. I'd be interested in anyone has a workaround.