0
votes

I have a elastic cluster with 2 nodes with each configured to act as data node as well as master node.

Cluster Configuration:

index.number_of_shards: 5
index.number_of_replicas: 0

The shards are distributed across the 2 nodes as follows:

Node A : 0 2 3
Node B : 1 4

On Node B,shard 1 is stuck in initialization phase because it is running low on RAM.

I need to take down Node B so that i can increase the RAM and rejoin the cluster.So,do I need to move all shards from Node B to Node A before restarting the node? Or can I let the shards remain in the node and still restart the node?

1

1 Answers

1
votes

If you don't care about index availability during the restart period, you can restart node B. If you do care about availability, you can either:

  1. Ask ES to move shards from node B to node A via shard allocation filtering

  2. Create a replica which will copy over shards from node B to node A