Distributed Search

[Srijan <sh...@gmail.com>]

Hi All,

I am on Solr 8.9.0 and have multiple collections hosting data from
different sources. And I have two requirements, one to search against a
specified collection and the other to perform a distributed search across
all collections.

The point I am debating is if I want to go the SolrCloud route or if I can
still leverage the old way of performing distributed search using the
shards parameter. Is the old distributed search even supported in newer
Solr versions? Would you mind sharing your experience regarding distributed
search if you have implemented one? What is your environment like and any
pain points you might have encountered while supporting distributed search.

Thank you very much,
Srijan

Re: Distributed Search

[Shawn Heisey <ap...@elyograg.org>]

On 2/28/22 13:27, Srijan wrote:
> The point I am debating is if I want to go the SolrCloud route or if I can
> still leverage the old way of performing distributed search using the
> shards parameter. Is the old distributed search even supported in newer
> Solr versions? Would you mind sharing your experience regarding distributed
> search if you have implemented one? What is your environment like and any
> pain points you might have encountered while supporting distributed search.

SolrCloud makes a lot of things easier.  Because of the need for 
zookeeper, initial setup can have challenges that are not present if 
using standalone mode, but it's not super hard to set up.

When I first set up a distributed index on Solr 1.4.0, SolrCloud was 
just an idea that people were toying with ... it didn't land in an 
official release until version 4.0.  So I figured out how to distribute 
data among multiple shards at index time and how to set up a special 
core/handler to do distributed queries.  SolrCloud eliminates the need 
to do any of that.  I no longer work for the company with that 
deployment.  When I left it, it was running some 4.x release, 4.7 if I 
remember right, and it had not been migrated to cloud mode.  If I was 
still working with that index, I would be taking steps to get it 
upgraded to a modern version, but I would likely NOT be switching to 
cloud mode yet.

Solr still supports doing all of it yourself without Cloud mode for 
typical use cases.  A word of warning -- when entirely new features are 
implemented, they often target SolrCloud first, and using that feature 
on a server in standalone mode is not always guaranteed.  At some point 
in the future, Solr might ALWAYS run in cloud mode, but that decision 
has not been made yet.  Standalone mode is still very useful for me, and 
I think for many others.

Thanks,
Shawn