I am working on alfresco 4.0.e data. When doing a full lucene reindex I get some NPEs in Lucene, it gets to 10% and then nothing happens. CPU use drops to very low levels, alfresco doesn't start.
After some trawling, I found this (http://forums.alfresco.com/comment/67704#comment-67704) and thought I'd give it a try. So I removed all entries in alf_node with acl_id null. That required removing entries from alf_child_assoc where parent_node_id and child_node_id were in the set of alf_node entries with null acl_id.
An alfresco restart now still shows the Lucene NPEs but it continues to 100%. I don't see the "Orphan child node has been re-homed under lost_found" log entries (as I'd expect to see in post 4.0 due to: https://issues.alfresco.com/jira/browse/MNT-6023).
Was this a sane thing to do?
Thanks for any insight into how sane/insane this "solution" is.
It turns out this solves my reindexing issue. Only 2 entries needed to be removed from alf_child_assoc. A further upgrade to 4.2.d showed 2 orphan nodes moved to lost_found (likely the same nodes I removed from alf_child_assoc). And alfresco started in 4.0.e and also in 4.2.d.
I do have backups (working on a copy), so I can easily roll back if something I try fails.