Questions about typical/simple clustered Solr software and hardware architecture

[ken edward <ke...@gmail.com>]

Hello,

I am brand new to Solr, and trying to ramp up quick. Please correct me
if I am wrong, but from what I read, in a true production environment,
is it true that :

1. Solr is made up of only "node" processes and "zookeeper" processes?

2. Each node and zookeeper process ideally runs on it's own physical server?

3. Searches can be sent to any of the node processes?

4. A typical HA configuration would put a proxy or load balancer out
in front of the nodes to distribute the work?

Ken

Re: Questions about typical/simple clustered Solr software and hardware architecture

[Toke Eskildsen <to...@kb.dk>]

ken edward <ke...@gmail.com> wrote:
> 2. Each node and zookeeper process ideally runs on it's own physical server?

ZooKeepers, yes. Nodes, it depends.

One reason to run multiple nodes on a single physical machine is to avoid large JVM heaps. Large JVM heaps can be problematic as major stop-the-world garbage collections can take a long time.

We have a setup with 25 nodes per physical server, each with 8GB of heap. Running that as a single node per physical machine would mean ~200GB heap. I am sure it is possible to wrangle such a beast, but I'd rather spend my energy on Solr instead.

- Toke Eskildsen