First, if master/slave suits our use-case, there's no reason to go to SolrCloud. However, the following are some of the things you get with SolrCloud:
1> automatic document routing (irrelevant if you don't have enough docs to need more than one shard) 2> automatic fail-over/recovery if nodes go down, without having to implement re-indexing strategies for docs that might have been indexed to the master but not yet replicated to the slaves. 3> Near Real Time (NRT) searches on all the nodes. In master/slave setups you can't search a doc until it's been both indexed, then replicated which is probably many minutes. 4> Centralized configuration management via Zookeeper. 5+> there's lots more, but these are the high points. None of these are relevant in all situations. For instance, say you have an application that re-indexes your entire corpus once a day and your corpus fits on a single shard, there's no compelling reason to go to SolrCloud. In fact, there's added complexity you can just avoid. Adding capacity is just adding slaves as you indicated. Contrast that with a large collection spanning, say 10 shards that is continuously getting updated, add the requirement is that the new documents be searchable within 10 seconds of being received by the system-of-record. Add that nodes may come up and go down for some reason.... you get the idea ;). SolrCloud is really not about performance measured by query response time as much as management of large, multi-sharded collections. Best, Erick On Sun, Feb 15, 2015 at 9:05 PM, CKReddy Bhimavarapu <chaitu...@gmail.com> wrote: > Dear all, > > I don't get when and where we should use solr cloud so that we will get > best performance. How to know the peak point of solr master server? > > my question is when a solr can't handle the requests i.e if we reach the > peak point at that time if I increase my master configuration does that > slove the problem. > > thanks in advance, > ckreddybh. <chaitu...@gmail.com>
