#of shards per index by keeping the future growth of cluster in mind?


(Tarun Jangra) #1

I am looking for feedback of my conclusion. I observed number of articles
and discussions about defining number of shards per index.

*. By default Es is providing 5 shards and 1 replica for each index. I
believe, There is no particular formula to define number of shards per
index. It depends on the type of the data you are going to index in shards.
*. #of nodes >= #of shards * (#of replicas + 1) for each index.
*. For stats type of data. It is good to have 1 shards and 1 replica.
Because such type of data suppose to have so many documents but smaller
size and most probably having numbers only.

Thank you.


(Shay Banon) #2

You do want to have more shards than nodes, so you can scale out easily. As
for stats type of data, a single shard assumption is not really relevant,
even small documents can amount to a lot of data.

On Sun, Jun 24, 2012 at 9:33 PM, Rimpy tarun@izap.in wrote:

I am looking for feedback of my conclusion. I observed number of articles
and discussions about defining number of shards per index.

*. By default Es is providing 5 shards and 1 replica for each index. I
believe, There is no particular formula to define number of shards per
index. It depends on the type of the data you are going to index in shards.
*. #of nodes >= #of shards * (#of replicas + 1) for each index.
*. For stats type of data. It is good to have 1 shards and 1 replica.
Because such type of data suppose to have so many documents but smaller
size and most probably having numbers only.

Thank you.


(system) #3