Brian Moseley wrote:
i ran into a situation today where it looked like the lucene index
stopped being updated until i bounced my server, closing down and then
reopening the repository.
i observed some thread death stack traces in the server log right
around the time people were uploading resources to the server. they
appeared to be happening during indexing. i bounced the server and
didn't see any more thread deaths, but when we ran some queries, we
got zero results even when we could browse through the repository and
observe nodes there.
can you post those stack traces? that will probably help to find out
what could have caused the problem.
i thought of deleting and recreating the relevant nodes, but that
didn't have any effect. it was as if the index was not being updated,
even though i'd bounced the server and given the jvm more memory.
if you see OutOfMemoryExceptions this might be due to:
http://thread.gmane.org/gmane.comp.apache.jackrabbit.devel/5026
later in the day i bounced the server again, then deleted and
recreated the nodes being queried. suddenly the queries started
returning the expected results. i'm not sure what i might have done
differently this time.
anyway, does this sort of behavior ring any bells for anyone? maybe
something to do with merging indexes or persisting the volatile index
or something?
also, is there any builtin way to cause the index to be flushed and
rebuilt? seems like that would come in handy should this situation
occur again before we understand what's going on.
there is a configuration parameter that controls how long the volatile
part of the index is held in memory: volatileIdleTime
per default the volatile index is written to disk after 3 seconds.
if you want to rebuild the index you have to shutdown jackrabbit and
delete the index manually and then restart. it is not possible to
re-index a workspace while it is in use.
regards
marcel
