We have a 3node elasticsearch cluster with both master/data running fine now. Each node has 1TB SAN storage and occupied 2.5TB data and indexes are being created with 5 primary shards and 1 replica distributing among 3 nodes.
My query is what happens, if 1node is failed? Does Elasticsearch promote any replica as primary and adjusts data of node1 on two nodes? In that case my two nodes wouldn't accommodate 800GB. Or, data on node1 will be ignored and new indexes will be created on available two nodes with same 5 primary and 1 replica each?
Please correct me, if am wrong.
Does Elasticsearch promote any replica as primary
adjusts data of node1 on two nodes? In that case my two nodes wouldn't accommodate 800GB.
If disk usage is above the low disk watermark (https://www.elastic.co/guide/en/elasticsearch/reference/current/disk-allocator.html), Elasticsearch will not create a replica in order to not make the node run out of space. As a result of this your cluster will likely be in a yellow state since not all shards would be allocated.
Thank you for update. If node1 is down, how other nodes communicate with failed node to transfer data? And, if Elasticsearch promotes replicas as primary, what is the essential to transfer data from failed node since the data available on remaining nodes?
This topic was automatically closed 28 days after the last reply. New replies are no longer allowed.
It doesn't but you would typically have the data on some other node.
It would transfer the data from this new primary to a new node so that the number of replicas is still honored.