optimizing away the expensive cases is your best bet if you can do it ... another option is to use a custom HitCollector which keeps track of how long it's been running and throws a subclass of RUntimeException which you explicitly catch and deal with as appropriate if the query has been taking too long.
: Date: Wed, 21 Feb 2007 09:34:36 -0500 : From: Tim Johnson <[EMAIL PROTECTED]> : Reply-To: java-user@lucene.apache.org : To: java-user@lucene.apache.org : Subject: Stop long running queries : : I'm having issues with some queries taking in excess of 500 secs to run to : completion. The system being used consists of ~100 million docs spilt up : across ~600 indexes. The indexes are of various sizes from 15MB to 8GB and : all searches done in the system require an exact count of matching hits. : The documents are also indexed with various fields and types ranging from 15 : to 200 fields depending on the data indexed. : : 98% of all queries run under 4 sec, but the remaining 2% run for > 15 secs : and sometimes > 500 secs. These long running queries usually consist of : wildcard and range queries where the number of BooleanQuery statements : reaches into the 100,000 range. Some of these queries return a relatively : small number of results (below 300). : : I've recently tried to kill the search using a separate thread to wrap the : searcher.search method, which works, but only once. The next query to come : through throws an "IOException: The handle is invalid." in the : FSInputStream.readInternal() method. : : Any suggestions on how to stop these long running queries? : : Tim Johnson : SAIC : Systems Engineer : 757.226.1569 : : : : --------------------------------------------------------------------- : To unsubscribe, e-mail: [EMAIL PROTECTED] : For additional commands, e-mail: [EMAIL PROTECTED] : -Hoss --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
