SolrCloud auto routes the documents to the correct shard leader, however you would be able to reduce the extra hop by sending the document to the correct shard. Here are a few posts that explain how the document routing in SolrCloud works:
https://lucidworks.com/2013/06/13/solr-cloud-document-routing/ https://lucidworks.com/2014/01/06/multi-level-composite-id-routing-solrcloud/ If the extra hop isn't something you are much bothered about, I wouldn't suggest adding the complexity to your client code. The Java client that Solr ships with, SolrJ has a 'smart' client i.e. CloudSolrClient, that tracks the clusterstate in zk by keeping a watch. It also contains caching logic, and more to optimize sending requests to a SolrCloud cluster. You might want to explore that and possibly use that instead. -Anshum On Sun, Jan 29, 2017 at 9:49 AM GW <thegeofo...@gmail.com> wrote: > Hi folks, > > 1: Can someone point me to some good documentation on how this works? Or is > it so simple that I'm over thinking? > > My understanding of document routing is that I might be able to check the > hash of a shard with the hash of the document id and determine the exact > node / document and know exactly where to send the request reducing > Zookeeper traffic. > > I'm getting ready to deploy and I have used the recommended format in my > doc id > > All my work is REST/curl -> Solrcloud > > I plan to watch cluster status through the admin console REST to and build > a list of OK servers to do the reads for the website. > > I have a crawler that will be running mostly 3:am Eastern to 3:am Pacific, > outside the bulk of read activity. I plan to do all posts to Who Has > Zookeeper according the admin REST API > > Can I get some reassurance? Be gentle, this is my very first solrcloud > deployment and it's going to production. I'm about to write script for > something that I still feel I am week in concept. > > When I'm done and I totally understand, I promise to publish a nice A - Z > REST deployment HowTo for HA with class examples in (PHP,Perl,Python)/curl. > > > Best regards, > > GW >
