> ---------- Forwarded message ---------- > From: Ryan McKinley <[EMAIL PROTECTED]> > From the lucene 2.4 draft announcement... > > * IndexReader can be opened with new readOnly=true mode, which gives > better performance in a multi-threaded environment. > > I don't normally delve into lucene innards, so this may already be addressed. > Would it be worthwhile adding a 'readonly' mode for slave servers?
This patch already takes care of it: 10. SOLR-730: Use read-only IndexReaders that don't synchronize isDeleted(). This will speed up function queries and *:* queries as well as improve their scalability on multi-CPU systems. (Mark Miller via yonik) [...] > It looks like we have access to the read only searcher, but never call > it. From SolrCore.java: > // gets a non-caching searcher > public SolrIndexSearcher newSearcher(String name) throws IOException { > return newSearcher(name, false); > } This call was only used by the update handler to do get a searcher to do deletions... but we no longer do that (DUH2.openSearcher() is no longer called). So I think everything should be read-only. -Yonik