This is an impossibly vague question to answer. You haven't provided _anything_ useful to help answer. How much physical memory is on the machine? How man processors? What kind of searching? Faceting? Indexing rate? How many documents? What kind of documents?
What evidence do you have that you _need_ to run multiple JVMs? Is this a case of premature optimization? About the only reason to worry about this is if you are encountering long stop-the-world JVM pauses. Have you any evidence of that? Unless and until you do there's not much point in running multiple JVMs. Or, I guess, if your queries are taking too long to come back even under light load. And even with all that information, the only honest answer is "try some variations and find out". Here's a straw-man proposal: Start by allocating 16G to your JVMs until you have allocated a total of around 30% of your physical memory to your JVMs. Say you have 128G, that would be 2 JVMs. Then measure/tune/measure, _then_ decide. Best, Erick On Tue, Oct 4, 2016 at 7:55 PM, shalu <shalusingh0...@gmail.com> wrote: > We are also require multiple solr instances to divide the heap space between > cores. What is the best way to do it? > > > > -- > View this message in context: > http://lucene.472066.n3.nabble.com/Multiple-solr-instances-on-one-server-tp4248411p4299645.html > Sent from the Solr - User mailing list archive at Nabble.com.
