I want to use different storage options by my indices. Recent indices should stay on fast SSDs while older indices should be moved to classical HDDs to save storage costs. The hot / warm architecture looks quite like that what I need.
Can one node serve hot and warm patterns in parallel, deciding on some index tag (or whatever) on which storage to store, or do I need separate nodes for hot and warm storage?
My resources are limited to 3 physical hosts and I want to have as low overhead as possible.
If I need separate nodes I think of going to have 3 hot nodes (SSD) and 2 warm nodes (HDD), while the warm nodes are running on the same machines, just pointing to another path.data.
Is that approach correct? Any tipps or pitfalls I should be aware of?
Thanks a lot,