Primary Index allocation strategy


(Thibaut Britz) #1

Hi,

We have a cluster of multiple machines and would like to balance this
cluster by:

  1. Making sure that each of the nodes contains about the same number of
    primary indexes (_> query primary_first)
  2. We have multiple datadirectories on these nodes (each on a seperate hd).
    We would like the primary indexes to be split evenly on theses hds on one
    node.

Is there any way to achieve this? How are the primary indexes being chosen
on cluster start? On which criteria? Can we overwrite to which data
directory the index will be placed?

Thanks,
Thibaut

--


(BillyEm) #2

Thibaut: I bet this thread, once you get through it (and its worth it) will
answer yoour question.

http://elasticsearch-users.115913.n3.nabble.com/Multimaster-td3621103.html

On Monday, September 24, 2012 8:40:04 AM UTC-4, Thibaut wrote:

Hi,

We have a cluster of multiple machines and would like to balance this
cluster by:

  1. Making sure that each of the nodes contains about the same number of
    primary indexes (_> query primary_first)
  2. We have multiple datadirectories on these nodes (each on a seperate
    hd). We would like the primary indexes to be split evenly on theses hds on
    one node.

Is there any way to achieve this? How are the primary indexes being chosen
on cluster start? On which criteria? Can we overwrite to which data
directory the index will be placed?

Thanks,
Thibaut

--


(Thibaut Britz) #3

Thanks,

I also saw that there are some features in this direction planned for 0.20.
Looking forward to this :slight_smile:

Thibaut

On Wed, Sep 26, 2012 at 3:18 AM, BillyEm wmartinusa@gmail.com wrote:

Thibaut: I bet this thread, once you get through it (and its worth it)
will answer yoour question.

http://elasticsearch-users.115913.n3.nabble.com/Multimaster-td3621103.html

On Monday, September 24, 2012 8:40:04 AM UTC-4, Thibaut wrote:

Hi,

We have a cluster of multiple machines and would like to balance this
cluster by:

  1. Making sure that each of the nodes contains about the same number of
    primary indexes (_> query primary_first)
  2. We have multiple datadirectories on these nodes (each on a seperate
    hd). We would like the primary indexes to be split evenly on theses hds on
    one node.

Is there any way to achieve this? How are the primary indexes being
chosen on cluster start? On which criteria? Can we overwrite to which data
directory the index will be placed?

Thanks,
Thibaut

--

--


(system) #4