Routing & Nodes

Elastic Stack Elasticsearch
1

This is a very basic question, but I can't seem to find the answer in the
documentation (I'm sure I'm missing it somewhere). But regarding nodes &
request routing. If we have an index that is split up between several
nodes (lets say 10) each containing a shard, and now I want to make a
query, do I point my query to a master node which then sends the request to
each of the 10 nodes, compiles the result, and then sends the response back
to me? Or can I send the request to any of the 10 nodes, and it will do
the same (forward the request to the other 9 peers, compile and send back
the response)? Or is it some other flow entirely?

Thanks for your help,
-Ryan

--

2

Hello, send to master node : as es is a distributed system, master node
will do the job for you.
Le 16 janv. 2013 05:08, "ryano" ryanogle79@gmail.com a écrit :

This is a very basic question, but I can't seem to find the answer in the
documentation (I'm sure I'm missing it somewhere). But regarding nodes &
request routing. If we have an index that is split up between several
nodes (lets say 10) each containing a shard, and now I want to make a
query, do I point my query to a master node which then sends the request to
each of the 10 nodes, compiles the result, and then sends the response back
to me? Or can I send the request to any of the 10 nodes, and it will do
the same (forward the request to the other 9 peers, compile and send back
the response)? Or is it some other flow entirely?

Thanks for your help,
-Ryan

--

--

3

Thanks for your answer. A follow up question - if all queries are going
through a single master node, then no matter how much I shard to additional
worker nodes, there is a theoretical limit based upon how much throughput
the master can take, correct? What happens when my throughput exceeds what
the master can take?

--

4

On Wed, 2013-01-16 at 07:33 +0100, TheKnto wrote:

Hello, send to master node : as es is a distributed system, master
node will do the job for you.

No. You can send the request to any node. In fact you should round robin
through all the nodes to spread the load. Each node has the full
cluster info and knows how to scatter the request to the appropriate
shards, and to gather the result.

clint

--