Ok, more detail: I was testing using the NoMergePolicy in Solr. As Hoss pointed out in another thread, NoMergePolicy has no 0-argument constructor, and so throws an exception during loading the core.
When there is no existing data/index/ directory, Solr creates a new index/ directory at the beginning of loading the core, locks it, but does not flush out an empty index. Here's the problem: when the core fails while being loaded (in this case because the core configuration was bogus) it left the index/ directory locked. It did not flush out the new empty index (just the segment* files). So, if a core has no index, and fails during loading, it should either write out an empty index as it intended to, or remove the half-built data/index/ directory. Or just not make the empty index until loading completes? Lance On Wed, Nov 10, 2010 at 11:52 AM, Chris Hostetter <hossman_luc...@fucit.org> wrote: > > : There is now a data/index with a write lock file in it. I have not > : attempted to read the index, let alone add something to it. > : I start solr again, and it cannot open the index because of the write lock. > > Lance, i can't reproduce using trunk r1033664 on Linux w/ext4 -- what OS & > Filesystem are you using? > > If you load "http://localhost:8983/solr/admin/stats.jsp"; what does it list > for the "reader" and "readerDir" in the "searcher" entry? > > : Why is there a write lock file when I have not tried to index anything? > > No idea ... i don't get any write locks until i actually attempt to index > something. > > > > -Hoss > -- Lance Norskog goks...@gmail.com
