Quote:
Not on same machine
So
how is this index shared? are you using NFS? The use of NFS or other network filesystem to share the index is not recommended, look at the Lucene forums: you'll find lots of problems.
In Search each index write operation needs the lock, then it guarantees to release it after the operation, but there is no notion of a "fair lock across machines", so under load it is probably re-acquiring the lock right after, so node2 never happens to "see" the lock available and after many failures it times out.
You'll find more insight and many solutions in the book.