If nobody objects I am going to commit this in the next 24 hours.
Dennis On 07/19/2010 04:00 PM, Dennis Kubes (JIRA) wrote:
DistributedBeans should not close their RPC counterparts -------------------------------------------------------- Key: NUTCH-857 URL: https://issues.apache.org/jira/browse/NUTCH-857 Project: Nutch Issue Type: Bug Affects Versions: 1.1 Environment: All Reporter: Dennis Kubes Assignee: Dennis Kubes Fix For: 1.2 Attachments: NUTCH-857-1-20100619.patch DistributedSearch and Segment Beans currently call close on their RPC counterparts from their own close methods. This results in killing (closing) all distributed servers when the main bean (website, application, etc) is shutdown. DistributedSearchServer (SegmentServer) are run independent from the main NutchBean or website calling those servers in shard type environments. With the current code the distributed servers are closed and any further search requests throw IndexAlreadyClosed exceptions. The distributed servers have to be restarted before searching can resume. Obviously this doesn't work in a large distributed search where multiple beans could be called the distributed servers and where distributed servers could be coming up and down frequently. The solution is simple though. The Distributed beans shouldn't call close on their RPC counterparts. Patch is attached.