Bad performance of 2-node-cluster


(Herbert Bodner) #1

Hey,
I have about 7 million documents in my ES cluster using default
configurations (5 shards, 1 replica).
I set up 4 virtual machines with similar configurations on a 8 core server,
each virtual machine assigned with 2 virtual cores and 4 GB memory. On each
node I installed Elasticsearch.
I did several search tests to measure query response time and the results
are a bit confusing:

  1. When using only a single node, I get the best response time.
  2. When using a 2-node cluster response time increases significantly,
    meaning the response time gets worse (the more found documents the worse
    the response time)
  3. For a three- or four-node cluster the response time improves again and
    is at the level of the single-node configuration.

Any idea why the 2-node cluster has such a bad performance compared to the
other configurations?

cheers
Herbert

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/4068f0a2-50dc-4f4a-9f51-6e12dced4e80%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


(Jörg Prante) #2

Why do you set up more than one JVM per machine?

Jörg

On Wed, Mar 19, 2014 at 10:36 AM, Herbert Bodner
herbert.bodner@gmail.comwrote:

Hey,
I have about 7 million documents in my ES cluster using default
configurations (5 shards, 1 replica).
I set up 4 virtual machines with similar configurations on a 8 core
server, each virtual machine assigned with 2 virtual cores and 4 GB memory.
On each node I installed Elasticsearch.
I did several search tests to measure query response time and the results
are a bit confusing:

  1. When using only a single node, I get the best response time.
  2. When using a 2-node cluster response time increases significantly,
    meaning the response time gets worse (the more found documents the worse
    the response time)
  3. For a three- or four-node cluster the response time improves again and
    is at the level of the single-node configuration.

Any idea why the 2-node cluster has such a bad performance compared to the
other configurations?

cheers
Herbert

--
You received this message because you are subscribed to the Google Groups
"elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an
email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit
https://groups.google.com/d/msgid/elasticsearch/4068f0a2-50dc-4f4a-9f51-6e12dced4e80%40googlegroups.comhttps://groups.google.com/d/msgid/elasticsearch/4068f0a2-50dc-4f4a-9f51-6e12dced4e80%40googlegroups.com?utm_medium=email&utm_source=footer
.
For more options, visit https://groups.google.com/d/optout.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/CAKdsXoGF%3DhT5BbT8dR6nuBq29CPB75kjY-SDwNvSZDX9tV%3DjDA%40mail.gmail.com.
For more options, visit https://groups.google.com/d/optout.


(Herbert Bodner) #3

Why do you set up more than one JVM per machine?

I didn´t do that!? One Elasticsearch-Installation on each of the four
virtual machines.

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/0dd8ca64-4f29-4dcb-8c1c-cecaf2f35f3a%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


(Herbert Bodner) #4

I guess I found the problem: The results that 2 nodes perform much worse,
cannot be reproduced.
There has been going on something else on the server at that time when I
executed the tests for 2 nodes, which influenced the results. (Probably
somebody copied a large file to or from the server, which made the disk
kind of busy)

--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/67983dda-dcf0-4ef5-a566-8a7a00f51761%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.


(system) #5