Hello Elastic experts,
I would like to migrate from a VM based deployment to one on kubernetes - and eck. Because of the size of the indices (many TBs of data) it has been suggested that I can create data nodes and join the cluster setting discovery seed nodes to the existing masters. Then moving shards with a routing allocation strategy excluding all existing nodes.
Question 1. It appears that an ECK cluster must require at least one master node, so does this mean that if I start a cluster I will necessarily adding master nodes if each nodes that I define has both master and data?
Question 2. Recently discovered daily snapshots on the old cluster that are stored to Azure storage, which opens up the possibly of migrating from snapshot.
So should I pursue snapshot migration approach?
Hi
If you deploy Elasticsearch with ECK, the cluster must have at least one master eligible node. However, this does not mean you are forced to add new dedicated masters if your nodes already have both roles. If your nodes are configured as master and data, they can participate in master election as usual. The important point is to maintain a healthy quorum and avoid unstable master elections. In most production setups with ECK, it is still recommended to run dedicated master nodes to keep cluster coordination stable.
Given the data volume (many TBs), restoring from snapshot is usually the safest and simplest approach. Joining new nodes and moving shards across environments can work, but it adds operational complexity and risk during relocation. Since you already have snapshots stored in Azure, restoring them into the new ECK cluster is typically the cleaner migration path.