How to avoid yellow server status when updating without any indexing and querying


(Jae) #1

I've been running elasticsearch cluster for more than 2 months. It was a
long journey and I finally came to the kafka river, which can make
elasticseach cluster more resilient because data pulling model is much more
stable than pushing model. Good thing is, I can totally stop elasticsearch
indexing without any data loss.

Question is, even though I stopped all indexing by stopping all river
instances, when I restarted one elasticsearch instance, the server status
became yellow. So, how can I update elasticsearch cluster without making
the server status yellow even there's no indexing and querying?

Thank you
Best, Jae

--


(Igor Motov) #2

The yellow state means that at least one of the replicas is not yet
available. Let's say you have one replica per shard. When you shut down a
node, you remove all shards on this node from service. It means that
cluster losses at least some of the replicas and it needs some time to
rebalance and allocate these lost replicas somewhere else. The process
takes time and while cluster is doing it, it's state turns yellow. This is
normal and not something that you should worry about.

On Tuesday, November 6, 2012 8:26:07 PM UTC-5, Jae wrote:

I've been running elasticsearch cluster for more than 2 months. It was a
long journey and I finally came to the kafka river, which can make
elasticseach cluster more resilient because data pulling model is much more
stable than pushing model. Good thing is, I can totally stop elasticsearch
indexing without any data loss.

Question is, even though I stopped all indexing by stopping all river
instances, when I restarted one elasticsearch instance, the server status
became yellow. So, how can I update elasticsearch cluster without making
the server status yellow even there's no indexing and querying?

Thank you
Best, Jae

--


(system) #3