Lots of ES best practice articles recommend having dedicated master nodes.
Specifically, that would involve setting these flags:
node.master: true
node.data: false
Say, you had 7 index nodes and 3 master nodes
(https://blog.hipchat.com/2013/10/16/how-hipchat-scales-to-1-billion-messages/)
in your cluster. Only one of those three master-eligible nodes are going
to be elected master; what do the other two do? Do they just exist for
resiliency for when the elected master goes down? Do they literally not
have any other role when not elected?
As a followup, any ratio guidelines for indexing nodes vs dedicated
masters. From what I can tell, it wouldn't make much sense to run with
only one-dedicated master node; if that node goes down, your whole cluster
becomes unavailable.
On Monday, August 11, 2014 2:18:08 PM UTC-4, Matt Hughes wrote:
Lots of ES best practice articles recommend having dedicated master
nodes. Specifically, that would involve setting these flags:
node.master: true
node.data: false
Say, you had 7 index nodes and 3 master nodes ( https://blog.hipchat.com/2013/10/16/how-hipchat-scales-to-1-billion-messages/)
in your cluster. Only one of those three master-eligible nodes are going
to be elected master; what do the other two do? Do they just exist for
resiliency for when the elected master goes down? Do they literally not
have any other role when not elected?
They can act as search nodes, otherwise they don't do much.
There are no guidelines other than using an odd number of masters, and as
you pointed out, more than just 1. We currently have 3 masters for 25 data
nodes.
As a followup, any ratio guidelines for indexing nodes vs dedicated
masters. From what I can tell, it wouldn't make much sense to run with
only one-dedicated master node; if that node goes down, your whole cluster
becomes unavailable.
On Monday, August 11, 2014 2:18:08 PM UTC-4, Matt Hughes wrote:
Lots of ES best practice articles recommend having dedicated master
nodes. Specifically, that would involve setting these flags:
node.master: true
node.data: false
Say, you had 7 index nodes and 3 master nodes (https://blog.hipchat.com/
2013/10/16/how-hipchat-scales-to-1-billion-messages/) in your cluster.
Only one of those three master-eligible nodes are going to be elected
master; what do the other two do? Do they just exist for resiliency for
when the elected master goes down? Do they literally not have any other
role when not elected?
The HTTP module is not disabled so the three master eligible nodes can also
serve as result servers. Result servers can collect shard responses from
other nodes for sending the merge result back to clients, without having to
search or index. This is similar to a kind of a bridge or a proxy server.
Lots of ES best practice articles recommend having dedicated master
nodes. Specifically, that would involve setting these flags:
node.master: true
node.data: false
Say, you had 7 index nodes and 3 master nodes ( https://blog.hipchat.com/2013/10/16/how-hipchat-scales-to-1-billion-messages/)
in your cluster. Only one of those three master-eligible nodes are going
to be elected master; what do the other two do? Do they just exist for
resiliency for when the elected master goes down? Do they literally not
have any other role when not elected?
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.