Re: Solr irregularly having QTime > 50000ms, stracing solr cures the problem

2014-07-12 Thread IJ
GUess - I had the same issues as you. Was resolved http://lucene.472066.n3.nabble.com/Slow-QTimes-5-seconds-for-Small-sized-Collections-td4143681.html was resolved by adding an explicit host mapping entry on /etc/hosts for inter node solr communication - thereby bypassing DNS Lookups. -- View

Re: Group only top 50 results not All results.

2014-07-12 Thread Joel Bernstein
I agree with Alex a PostFilter would work. But it would be a somewhat tricky PostFilter to write. You would need to collect the top 50 documents using a priority queue in the DelegatingCollector.collect() method. Then in the DelegatingCollector.finish() method you would send the top documents to th

Re: Group only top 50 results not All results.

2014-07-12 Thread Alexandre Rafalovitch
I don't think either grouping or faceting work as postfilter. Otherwise, that would be one way. Have a custom post-filter that only allows top 50 documents and have grouping run as an even-higher-cost postfilter after that. Regards, Alex. Personal: http://www.outerthoughts.com/ and @arafalov So

Re: Group only top 50 results not All results.

2014-07-12 Thread Erick Erickson
You could also return the top 50 groups. That will certainly contain the top 50 responses. The app layer could then do some local sorting to figure out what was correct. Maybe you'd be returning 3 docs in each or something... I'd probably only go there if Michael's approach didn't work out though.

Re: Changing default behavior of solr for overwrite the whole document on uniquekey duplication

2014-07-12 Thread Erick Erickson
bq: But does performance remain same in this situation No. Some documents will require two calls to be indexed. And you'll be sending one document at a time rather than batching them up. Of course it'll be slower. But will it still be "fast enough"? Only you can answer that. If it's _really_

Re: Solr 4.x and master-slave schema

2014-07-12 Thread Erick Erickson
Shouldn't be hard at all. The only trick will be putting the configuration files into ZooKeeper from the collection that exists already. Here's what I'd try (WARNING: this worked for me on a simple test, I didn't test it robustly) if you can't re-index: 1> copy your index somewhere for backup. Th

Re: Reference numbers for major page fauls per seconds, index size, query throughput

2014-07-12 Thread Erick Erickson
If the stats you're reporting are during the load test, your CPU is kind of idling along at < 20% which supports your theory. Just to cover all bases, when you bump the number of threads jmeter is firing does it make any difference? And how many rows are you returning? This latter is important bec