Yu Shen & Arian: We can't help much without more information. In particular, how are the fields in question analyzed? What is the result of looking at the admin/analysis page? What do you get when you attach &debugQuery=on to the query?
You might review: http://wiki.apache.org/solr/UsingMailingLists But at a wild guess, you have something like WordDelimiterFilterFactory in your analysis chain, and it's splitting up your input into "www" "someurl" "com" as separate tokens, and www matches all documents so Solr is having to score all documents in your corpus, but that's just a guess. See the admin/schema browser page and find the most frequent terms for the field in question, that should indicate whether you have some tokens that appear in all docs. Try searching on plain "someurl". Is that slow? Or "someurl.anotherpart" or whatever. Best Erick 2012/1/9 François Schiettecatte <fschietteca...@gmail.com>: > About the search 'referal_url:*www.someurl.com*', having a wildcard at the > start will cause a dictionary scan for every term you search on unless you > use ReversedWildcardFilterFactory. That could be the cause of your slowdown > if you are I/O bound, and even if you are CPU bound for that matter. > > François > > > On Jan 8, 2012, at 8:44 PM, yu shen wrote: > >> Hi, >> >> My solr document has up to 20 fields, containing data from product name, >> date, url etc. >> >> The volume of documents is around 1.5m. >> >> My symptom is when doing url search like [ url:*www.someurl.com* >> referal_url:*www.someurl.com* page_url:*www.someurl.com*] will get a >> extraordinary long response time, while search against all other fields, >> the response time will be normal. >> >> Can anyone share any insights on this? >> >> Spark >
