# Master nodes with no replica

**URL:** <https://discuss.elastic.co/t/master-nodes-with-no-replica/62965>\
**Category:** Elasticsearch\
**Created:** [October 13, 2016, 3:30pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965 "2016-10-13T15:30:01Z")\
**Posts on this page:** 7\
**Page:** 1

<div class="post-metadata">

**Author:** ![minniek](https://avatars.discourse-cdn.com/v4/letter/m/aeb1de/32.png) [@minniek](https://discuss.elastic.co/u/minniek)\
**Post date:** [October 13, 2016, 3:30pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/1 "2016-10-13T15:30:01Z")

</div>

I would like to know how master nodes works without replica.  
If I have an structure like I said before (2 nodes in different servers) and I need to execute a search/insert/delete/etc what node do I need to use in order to get the query?  
I understand that the data is shared across the two nodes but I am not sure how it works. Do I need to create an client node in order to merge the data (remember I dont have replica enabled)?

Thanks

---

<div class="post-metadata">

**Author:** ![magnusbaeck](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/magnusbaeck/32/44943_2.png) [@magnusbaeck](https://discuss.elastic.co/u/magnusbaeck)\
**Post date:** [October 13, 2016, 3:33pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/2 "2016-10-13T15:33:25Z")

</div>

> If I have an structure like I said before (2 nodes in different servers) and I need to execute a search/insert/delete/etc what node do I need to use in order to get the query?

You can connect to any node in the cluster.

> I understand that the data is shared across the two nodes but I am not sure how it works. Do I need to create an client node in order to merge the data (remember I dont have replica enabled)?

No. The ES node you connect to (again, either one) will merge the data for the connecting client.

---

<div class="post-metadata">

**Author:** ![minniek](https://avatars.discourse-cdn.com/v4/letter/m/aeb1de/32.png) [@minniek](https://discuss.elastic.co/u/minniek)\
**Post date:** [October 13, 2016, 3:44pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/3 "2016-10-13T15:44:48Z")

</div>

But If I execute so many petitions per seconds to a one node. Do you think this is the best architecture for that?

---

<div class="post-metadata">

**Author:** ![magnusbaeck](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/magnusbaeck/32/44943_2.png) [@magnusbaeck](https://discuss.elastic.co/u/magnusbaeck)\
**Post date:** [October 13, 2016, 4:29pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/4 "2016-10-13T16:29:18Z")

</div>

It's a good idea to round-robin requests between the nodes of the cluster. In some cases one might want to have client nodes that don't store data and just deal with queries and other requests, but that's typically not necessary or cost-efficient for small clusters.

---

<div class="post-metadata">

**Author:** ![minniek](https://avatars.discourse-cdn.com/v4/letter/m/aeb1de/32.png) [@minniek](https://discuss.elastic.co/u/minniek)\
**Post date:** [October 13, 2016, 4:47pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/5 "2016-10-13T16:47:47Z")

</div>

I have 2 servers (Memory RAM= 32GB , HardDisk = 7.2T) and 1 node per server.  
So, are you telling me It is better to have two master nodes without a client node because it cost a lot for small cluster?  
I want to make sure that I have the best architecture for my hardware

Thanks

---

<div class="post-metadata">

**Author:** ![magnusbaeck](https://sea2.discourse-cdn.com/elastic/user_avatar/discuss.elastic.co/magnusbaeck/32/44943_2.png) [@magnusbaeck](https://discuss.elastic.co/u/magnusbaeck)\
**Post date:** [October 13, 2016, 4:52pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/6 "2016-10-13T16:52:19Z")

</div>

If you have just one client node it becomes a single point of failure, so maybe you'd actually want more than one? But I suspect that money is better spent on getting a well-equipped cluster overall. Whether it's a good idea with dedicated client nodes also depends on the kind of queries you make, and how many of them.

---

<div class="post-metadata">

**Author:** ![system](https://us1.discourse-cdn.com/elastic/original/3X/1/a/1ac57faf039f6b580b3f104ef42a2a89e41014de.png) [@system](https://discuss.elastic.co/u/system)\
**Post date:** [July 5, 2017, 10:12pm UTC](https://discuss.elastic.co/t/master-nodes-with-no-replica/62965/7 "2017-07-05T22:12:35Z")

</div>


