We had an issue recently in which one of our web application using
Lucene 1.4-final started to max out CPU under heavy load with ~20
concurrent users during testing. Web application was not able to serve
any more pages after this and we could see all lucene threads "wating
for monitor" in the JVM thread dump.
Oddly this only happens on Solaris 10 box, with 1.5 JDK. Windows XP has
no issue in dealing with a concurrent access of more then 50 users with
compatible index size.
We have tried all possible options to fix the issue on Solaris, but have
not been successful so far. The problem is quiet similar to what has
been posted in the following thread.