rzezeski at gmail.com
Tue Mar 15 18:52:15 EDT 2011
No, your data doesn't need to reside on the filesystem before you can index
it. If you haven't already figured it out, riak-search is a superset of
riak-kv. That means that you have all the features of riak-kv along with
the ability to index your data. That means you really don't need to deploy
both. Although, if you want to do it that way it's certainly possible--I
just want to make sure you are clear on this fact.
One way to index data is to parse documents on the filesystem to be indexed
and stored by riak-search. Another way is to index data as it flows into a
bucket which seems to be what you are after. To do that you need to add a
pre-commit hook specific to riak-search. If you haven't already, please
take a look at
If you have specific questions after reading that I'll be glad to
On Tue, Mar 15, 2011 at 12:53 PM, Saurabh Sehgal <saurabh.r.s at gmail.com>wrote:
> Hi Sean,
> Please let me know if I am wrong - but according to my understanding, the
> documents to index should reside on disk (the example on the wiki indexes
> *.txt files in a directory). What if my data is not written to files, but
> exists in a riak bucket. Do I have to pull it out and store it in a file on
> disk to be able to index it ?
> Also, do you know if there are any erlang clients/helper libs for solr. I
> could not find any. Can the post commit hooks in riak be written in Java ?
> Thank you,
> On Tue, Mar 15, 2011 at 5:27 AM, Sean Cribbs <sean at basho.com> wrote:
>> If you want to go with Solr, you have a few options:
>> 1) Use application code to post an <add> document to Solr when saving your
>> data into Riak.
>> 2) Implement a Riak post-commit hook in Erlang to do the same, or to queue
>> the additions to another process that can submit them to Solr.
>> I think you'll find that while on-disk indexes sound bad, that's actually
>> how Solr does them too, and not every index query will hit disk on either
>> system. The primary difference is that Riak Search is tightly integrated
>> with Riak KV and so you get a lot for free when you choose Riak Search,
>> including auto-indexing and MapReduce queries that start with a search.
>> Sean Cribbs <sean at basho.com>
>> Developer Advocate
>> Basho Technologies, Inc.
>> On Mar 15, 2011, at 2:08 AM, Saurabh Sehgal wrote:
>> Since riak search is marked as "Beta Software" on the wiki, I am skeptical
>> to use it in production.
>> Does riak offer any APIs or integration points that would allow me to
>> index data into Solr ?
>> Specifically, I am storing key/value pairs in riak, where the value is a
>> binary erlang term containing a tuple of text I would like to index.
>> Also, going through the riak search documentation, the erlang API seems to
>> only have the capability of indexing documents
>> on disk as files (please correct me if this isn't true). Is there
>> functionality to automatically retrieve and index documents already stored
>> in riak ?
>> Saurabh Sehgal
>> E-mail: saurabh.r.s at gmail.com
>> Phone: <425-269-1324>425-269-1324
>> LinkedIn: http://www.linkedin.com/pub/1/7a3/436
>> riak-users mailing list
>> riak-users at lists.basho.com
> Saurabh Sehgal
> E-mail: saurabh.r.s at gmail.com
> Phone: 425-269-1324
> LinkedIn: http://www.linkedin.com/pub/1/7a3/436
> riak-users mailing list
> riak-users at lists.basho.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the riak-users