Problems using distributed More Like This

2013-02-11 Thread Shawn Heisey
SOLR-788 added Distributed MLT to Solr 4.1, but I have not been able to 
get it to work.  I don't know if it's user error, which of course is 
very possible.  If it is user error, I'd like to know what I'm doing 
wrong so I can fix it.  I am actually using a recent checkout of Solr 
4.2, not the released 4.1.


I put some extensive information on SOLR-4414, an issue filed by another 
user having a similar problem.  If you look for the last comment from me 
on Feb 7 that has a code block, you'll see Solr's response when I use 
MoreLikeThisComponent.


https://issues.apache.org/jira/browse/SOLR-4414

Only the last seven of the query parameters were included on the URL - 
the rest of them are in solrconfig.xml.  Due to echoParams=all, the only 
part of the request handler definition that you can't see in the 
response is the fact that last-components contains spellcheck.


I redacted the company domain name from the shards and the one document 
matching the query from the result tag, but there are no other changes 
to the response.


If I send an identical query to the shard core that actually contains 
the document rather than the core with the shards parameter, I get MLT 
results.


I have heard recently that Solr 4.x has hardcoded the unique field name 
for SolrCloud sharding as id ... but my uniqueKey field name is tag_id. 
 Could this be my problem?  It would be a monumental development effort 
to change that field name in our application.  I am not using SolrCloud 
for this index.


Thanks,
Shawn


Re: Problems using distributed More Like This

2013-02-11 Thread Mark Miller
Eventually, I'll get around to trying some more real world testing. Up till 
now, no dev seems to have a real interest in this. I have 0 need for it 
currently, so it's fairly low on my itch scale, but it's on my list anyhow.

- Mark

On Feb 11, 2013, at 12:26 PM, Shawn Heisey s...@elyograg.org wrote:

 SOLR-788 added Distributed MLT to Solr 4.1, but I have not been able to get 
 it to work.  I don't know if it's user error, which of course is very 
 possible.  If it is user error, I'd like to know what I'm doing wrong so I 
 can fix it.  I am actually using a recent checkout of Solr 4.2, not the 
 released 4.1.
 
 I put some extensive information on SOLR-4414, an issue filed by another user 
 having a similar problem.  If you look for the last comment from me on Feb 7 
 that has a code block, you'll see Solr's response when I use 
 MoreLikeThisComponent.
 
 https://issues.apache.org/jira/browse/SOLR-4414
 
 Only the last seven of the query parameters were included on the URL - the 
 rest of them are in solrconfig.xml.  Due to echoParams=all, the only part of 
 the request handler definition that you can't see in the response is the fact 
 that last-components contains spellcheck.
 
 I redacted the company domain name from the shards and the one document 
 matching the query from the result tag, but there are no other changes to 
 the response.
 
 If I send an identical query to the shard core that actually contains the 
 document rather than the core with the shards parameter, I get MLT results.
 
 I have heard recently that Solr 4.x has hardcoded the unique field name for 
 SolrCloud sharding as id ... but my uniqueKey field name is tag_id.  Could 
 this be my problem?  It would be a monumental development effort to change 
 that field name in our application.  I am not using SolrCloud for this index.
 
 Thanks,
 Shawn