You are using a local filesystem and not like NFS? If you are not replicating, I'm surprised this would happen and doubt it's the bug mentioned in the other reply.
It could just be a bug we have to defend against - if a file is not there because the index is changing as we are counting, we should probably just continue on. - Mark On Nov 30, 2012, at 4:15 PM, Shawn Heisey <s...@elyograg.org> wrote: > This is branch_4x, checked out 2012-11-28. Here is my solr log, created by > log4j at WARN level: > > http://dl.dropbox.com/u/97770508/solr-2012-11-30.log > > There are a bunch of unusual exceptions in here. Most of them appear to be > related to getting information from the mbeans handler, specifically getting > the size of the index from the replication handler ... it appears that for > this exception, the index gets updated while the stats are being gathered, > and files in the index disappear. > > The other exceptions look like they are I/O related. I don't know why they > are happening. Can anyone shed any light on the situation? What other > information should I include for troubleshooting? > > For the disappearing index files, I am guessing that some kind of > synchronization may need to be added so that the index cannot change while > the replication handler is gathering the size information. I am actually not > using replication, the only reason I have the handler configured is so that I > can get the size of my indexes. See SOLR-3990. > > https://issues.apache.org/jira/browse/SOLR-3990 > > I think I need to file jira issues for these problems, but I would like to > understand them better before doing that. Can anyone offer any insight? > > Thanks, > Shawn >
