Clustering across co-locations


(Craig Warren) #1

Hi,

I'm currently evaluating ElasticSearch to see if it fits our needs. One
major requirement is to have our index available in 2
geographic co-locations, but also minimize the internet traffic between
these locations.

I want to be able to have ES clients in each location execute their queries
only on the nodes in the local network. Is there a way to configure the
cluster for this? This would require co-location's node set having each
shard or it's replica, and restricting where the query can run.

Is this possible to configure, and if so, how?

Thanks!

--


(Radu Gheorghe) #2

Hi Craig,

I think what you can do here is to split your data in two indices - one per
location - although you can easily search in both if you want to. Then,
keep each index within the limits of its location by Shard Allocation
settings:

http://www.elasticsearch.org/guide/reference/index-modules/allocation.html

After doing that, data shouldn't migrate between locations, and you could
enable clients to search only in one of the two indices.

On Saturday, September 1, 2012 2:42:45 AM UTC+3, Craig Warren wrote:

Hi,

I'm currently evaluating ElasticSearch to see if it fits our needs. One
major requirement is to have our index available in 2
geographic co-locations, but also minimize the internet traffic between
these locations.

I want to be able to have ES clients in each location execute their
queries only on the nodes in the local network. Is there a way to configure
the cluster for this? This would require co-location's node set having each
shard or it's replica, and restricting where the query can run.

Is this possible to configure, and if so, how?

Thanks!

--


(Craig Warren) #3

Thanks Radu, that looks like exactly what I needed.

On Sunday, 2 September 2012 22:51:58 UTC-7, Radu Gheorghe wrote:

Hi Craig,

I think what you can do here is to split your data in two indices - one
per location - although you can easily search in both if you want to. Then,
keep each index within the limits of its location by Shard Allocation
settings:

http://www.elasticsearch.org/guide/reference/index-modules/allocation.html

After doing that, data shouldn't migrate between locations, and you could
enable clients to search only in one of the two indices.

On Saturday, September 1, 2012 2:42:45 AM UTC+3, Craig Warren wrote:

Hi,

I'm currently evaluating ElasticSearch to see if it fits our needs. One
major requirement is to have our index available in 2
geographic co-locations, but also minimize the internet traffic between
these locations.

I want to be able to have ES clients in each location execute their
queries only on the nodes in the local network. Is there a way to configure
the cluster for this? This would require co-location's node set having each
shard or it's replica, and restricting where the query can run.

Is this possible to configure, and if so, how?

Thanks!

--


(system) #4