Sum of shards on one node


(Arun A Nayagam) #1

Hi,

I have a 6.3.0 - 3 node cluster collecting logs with default number of shards and replicas.
Each node is associated with a 300GB external disk.
Is it by design that one node will have the sum of shards? I would have expected it to distribute equally among the nodes.
The issue is when the max shard node reaches the file limit, ES starts to mark the indices as read only and things start to go ugly.
With this approach how can I use all the 900GB available?
Is there a way to allocate shards equally?
Here's a screenshot of that setup.

Any help is appreciated.

Thanks,
Arun


(Arun A Nayagam) #2

Hi,

I figured it out with the help of a colleague of mine.
Basically we had these running in AWS and since one of the node was in Zone B and the other two were in Zone A.
And since we had the below setting on,

cluster.routing.allocation.awareness.attributes: aws_availability_zone

It replicates everything from node1 and node2 to node3.

Our solution doesn't require this level of reliability.
If Zone B goes down its not a big deal.
So we switched it off like this,

{{server}}/_cluster/settings
    {
        "persistent": {
        	"cluster.routing.allocation.awareness.attributes": ""
      }
    }

And the reallocation process ensured all three had more or less the same number of shards.

But it's a really good setting to have it on for critical data.

Thanks,
Arun


(system) #3

This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.