The NFS mount has to be done with distributed file locking. I don't know what DFL features are available.
OS Native file locking is the default in solrconfig.xml, and I think this should be used in your scenario. But doing this over NFS is not likely to work well. On Tue, Apr 6, 2010 at 6:42 AM, Paolo Castagna <castagna.li...@googlemail.com> wrote: > Hi, > I have an existing web application which is using Lucene (v2.1.0 and/or > v2.4.x) and which I'd like to gradually migrate to Solr. > > I am already using multiple cores, master/slave replication and SolrJ > to re-implement current functionalities. > > One use case I have is: backup/restore indexes. > > I am thinking to use another Solr master (to which I will not submit > updates) to expose current Lucene indexes (on NFS) and allow them to be > replicated to my real Solr master (which I am using to submit updates > to). This way, I can reuse the current restore capabilities I have and > propagate "restored" indexes to my Solr cluster. To clarify: > > +-------------+ +--------------+ +--------------+ > | Solr master | | Solr master | |+--------------+ > | (read-only) |<-(r)--| (read/write) |<-(r)--|| Slaves | > +-------------+ +--------------+ +| | > ^ ^ +--------------+ > | | > | | > +-----+ updates > | NFS |<---+ via SolrJ > +-----+ | > ^ | > | +--- updates via legacy Lucene > restore > > (r) = replicates from > > Before trying to test this setup, I'd like to have a feedback and see > if there are issues with it and/or better alternatives. > > Aslo, I am not 100% sure I can have Solr v1.4 running concurrently with > other applications using Lucene (v2.1.0 and/or 2.4.x) pointing at the same > index (on an NFS mounted file system). > > Is <lockType>simple</lockType> the recommended setting in this scenario? > > Paolo > -- Lance Norskog goks...@gmail.com
