0

How to scale out my solr 3.5 installation? How to replicate indexes? How to configure load balancing for POST and GET operations? How to auto-scale the servers? When load is low, put some servers down. When load is high, automatically add new servers to the cluster

Daniel Cukier
  • 11,502
  • 15
  • 68
  • 123
  • 1
    is there a reason why you're not looking at solr4.0? Also, for removing servers, you might want to look at Amazon AWS or a cloud based solution (if you're looking to host on your own) that allows for autoscaling such as Heat/Reactor with OpenStack. – ali haider Apr 14 '14 at 01:30
  • I can't upgrade to solr 3.5 now, because there are some customized features that work only on 3.5 and need to be re-written before migrating to 4.0 :-( – Daniel Cukier Apr 14 '14 at 01:31
  • for load balancing, you can use HA Proxy/Varnish/Nginx if you're only looking to load balance based on post/get calls. Not sure about solr 3.5 - I know I saw a link where somebody made changes to solr3.5 but I have not used it. I don't think putting a distributed data grid like hazelcast/inifinispan in front of solr would work for most cases (you probably have a better idea of how your data should be indexed) – ali haider Apr 14 '14 at 01:50

1 Answers1

0

For solr scalability, you can use data replication as described in this page. One node will be the master and the others will be slave. Put everybody behind any load balancing software for GET operations. You will have to take care that you do the POST operations directly to the master server (don't use the load balancing for that). The data will be automatically replicated to the slaves.

Auto-scaling depend on the cloud provider your are using. From what I know, Amazon AWS and Rackspace have solutions, web configurable. Basically, you will need to create an image of the slave server and use this image as the core image for the auto-scaling. Then, all you need is to configure auto-scaling policies.

Daniel Cukier
  • 11,502
  • 15
  • 68
  • 123