[jira] [Created] (SOLR-12363) Duplicates with random search, cursors, and fixed seed

Previous Topic Next Topic
 
classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

[jira] [Created] (SOLR-12363) Duplicates with random search, cursors, and fixed seed

JIRA jira@apache.org
Alexander S. created SOLR-12363:
-----------------------------------

             Summary: Duplicates with random search, cursors, and fixed seed
                 Key: SOLR-12363
                 URL: https://issues.apache.org/jira/browse/SOLR-12363
             Project: Solr
          Issue Type: Bug
      Security Level: Public (Default Security Level. Issues are Public)
    Affects Versions: 5.3.1
            Reporter: Alexander S.
         Attachments: Screen shot 2018-05-16 at 14.51.19.png, Screen shot 2018-05-16 at 14.51.23.png, Screen shot 2018-05-16 at 14.51.26.png

We do have a SolrCloud cluster and just updated one of our views to use cursors with the random order. Our goal was to use an infinite scroll with the random ordering so we can shuffle results once every 24 hours.

To do so we save the seed that we use in our random order to the cookies with the 24 hours expiration period, which didn't work as expected:
 # Results are shuffled with every request (every time we pass the initial cursor value "*" and the same random value for ordering we already used).
 # Results contain duplicates sometimes. Not a lot of them, but from time to time they appear.

In our *schema.xml* we have:
{code:java}
<fieldType name="rand" class="solr.RandomSortField" omitNorms="true"/>
<dynamicField name="random_*" stored="false" type="rand" multiValued="false" indexed="true"/>{code}
In our search requests, we order by *random_123 asc, id asc*, where *123* is the seed from cookies.

Here is the page [https://awards.wegohealth.com/nominees]

Even when I try to get the "next page" URL from google chrome developer console and open it in separate tabs it yields different results: [https://awards.wegohealth.com/nominees?cursor=AoJYmYbyATRBd2FyZDo6Tm9taW5lZSAxMzI0Mg%3D%3D]

So it feels like the seed parameter we use is ignored or every shard understands it differently, not sure.

On the screenshots, you can see the URL is the same and results are different.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: [hidden email]
For additional commands, e-mail: [hidden email]