Alfresco becomes unresponsive after bulkimport

[aweber1nj]

I am trying to test some small batches of documents to bulk-import (using the semi-official, semi-supported bulkimport tool).

Following the logs, even for just 20 documents (total) allocating 2 threads, the import seems to finish in under a minute (a little slow, but ok).  Because of a known issue with thumbnailing taking-up a lot of resources, we've actually disabled that temporarily.

However, very soon after the import completes, Alfresco (Share and Explorer, and any other way you try and access the repository) becomes virtually unresponsive, and Tomcat eats 100% of the CPU and just stays there.  I let it run for over an hour, and it just did not return to "normal"…even with only 20 documents added, and the logs seeming to indicate that the import was entirely finished after a few minutes.

EDIT: I have looked at catalina.out, alfresco.log and some others, but no errors appear other than ones that appear to be from when I was testing/trying Share during this issue and it was timing-out and failing.

I tried a jstack, but had to use the -F switch and it showed a lot of threads "BLOCKED", but not a lot of details about what was blocking them (but I would be happy to post those results if someone thinks they're worthwhile).

After restarting tomcat, CPU is pretty idle/normal again.  Tomcat using between 1 and 2%.

Has anyone else seen this and understand what is going on (and possibly how to workaround/fix it)?  I have a few thousand documents that I need to bulk load very soon, and I was really counting on this utility to do it for me.  Anyone have any tips as to how to isolate what COULD be causing the issue?

Thanks for any tips/tricks/advice.

(Alfresco 4.0.e on RHEL6 with postgresql)

-AJ

[aweber1nj]

I am continuing to search on this, and since I _do_ seem some indications about the AbstractFeedGenerator, I'm wondering if I'm running into the issue reported where some DB connections aren't being closed correctly or something?

Theoretically, a bulk import causes a lot of "changes" (New/Edited Documents) entries that, if I understand the Feed Generator correctly, would be reported to the users and it would therefore process those.

Anyone think it could be part of the problem, and have any ideas on why these DB connections are not being closed properly?  The JIRA entries seem to suggest that shutting down the generator may just move the issue elsewhere.  So is there a "bigger problem" somewhere that needs to be addressed?

If anyone has any ideas, I could really use the help.

Thanks,
AJ