+1
> Date: Mon, 15 Dec 2014 21:44:44 +1300 > Subject: Re: To understand SolrCloud configurations > From: esj.f...@gmail.com > To: solr-user@lucene.apache.org > > HI Shawn, > > Thanks, You have answered my question to a certain extend, But I wanted to > Isolate Solr Cloud from application and do some load testing by setting up > Jmeter Script. I could hit Solr instances, but it will not simulate how > Application (Client) will deal with Solr Cloud. Any suggestions for a > better way of achieving this? > > I want to do this, Solr keeps failing when we index large data and can't > find anything in the logs as well, I wanted to identify where it's failing, > I thought of using the above approach to make sure Solr and zookeeper setup > is correct , > Also, I would like to know a better way to debug Zookeeper and Solr, What I > have done so far is, > > 1. Make sure Zookeeper picking a Solr leader when the existing sun goes > down. > 2. Setup is working when one (lead) ZooKeeper is down etc ... > 3. Access Server runtime and see the data > > Thanks, > Shanaka > > On 15 December 2014 at 19:22, Shawn Heisey <apa...@elyograg.org> wrote: > > > > On 12/14/2014 12:41 PM, E S J wrote: > > > This question is related to the same configurations I've posted. How > > should > > > I manually test indexing via Zookeeper, I mean not directly accessing > > solr > > > nodes like, > > > curl http://solr1.internal:7083/solr/c-ins/update?commit=true -H > > > "Content-Type: text/xml" -d "@mem.xml" > > > > > > I have a solr client which uses CloudSolrServer to send request to > > > SolrCloud, But my intention is to isolate my SolrCloud and send index & > > > search requests to make sure Solr Cloud setup is working fine. Later I > > can > > > do Solrclient integration testing. How should I send index requests > > > manually ( like curl) to index data to solrcloud such a way > > CloudSolrServer > > > use ZooKeeper to LB/Pick Solr instance ? > > > > If you have either single-shard collections or multi-shard collections > > with automatic routing, SolrCloud is designed so that you can send any > > kind of request to any machine in the entire cloud, and it will be sent > > where it needs to go. If the collection uses manual (implicit) routing, > > then queries can go anywhere, but updates must be directed to the > > correct shard. > > > > If you are not using CloudSolrServer, then you must either set up a load > > balancer in front of SolrCloud, or your application will need to know > > where your Solr servers are. Curl cannot talk to zookeeper, because > > zookeeper does not speak HTTP. > > > > CloudSolrServer allows your application to specify only the zookeeper > > hosts, it doesn't need to know where the Solr servers are. This is > > because it includes a full zookeeper client. > > > > There is an API in Solr at /solr/zookeeper that can, with appropriate > > parameters, return various pieces of information from zookeeper in JSON > > format. This is the place where the admin UI gathers the information > > necessary to create the various options on the Cloud tab. Once your > > application has that information, it can use it to find out the Solr > > URLs to use. > > > > If this doesn't answer your question, please clarify it. > > > > Thanks, > > Shawn > > > >