Hi, I want to deploy elasticsearch on bare metal kubernetes.
I will use local storage as there is no fast and (by k8s) manageable storage provider available.
My k8s cluster consists of 5 servers (1 master, 4 workers), 2 of them have enough HDD capacity for elasitcsearch, the other two workers have only one small disk (only good for sstateless pods).
To learn for production I want to reach HA for my elastic stack, meaning I need at least 3 nodes. To ship around my disk issues, I have following idea, and ask you if that should work:
- server 1
- big disk, nod
- node.master: true
- node.data: true
- node.ingest: true
- server 2: big disk
- node.master: true
- node.data: true
- node.ingest: true
- server 3: low disk
- node.master: true
- node.data: false
- node.ingest: true
Does it make sense to run ingest on a non data node?
Would this scenery work?
Do you see any pitfalls?
I will work with 1 replica only.
Thanks, Andreas