I have a setup where I have some machines with SSD drives and others with slower spinning disks. I have some indices that I would like to place on the SSD machines so those requests are served more quickly. I have used allocation filtering thus far and it's worked well. However I am running out of space on the SSD drives and would still like to put more data there. I noticed there is a "preference" setting in searches that allow requests to prefer primary shards. So I was thinking that perhaps I could place only primary shards on the SSD machines and have the replicas on the spinning disks (other than disk space, the machines are not close to capacity at all). I looked through the documentation and I don't see any configuration settings to allow this. The closest thing I found was awareness, but that's not quite it: I could keep a full copy of the data would be on the SSD machines, but not necessarily the primary shards. And the search request "preference" also doesn't seem to allow for restricting the search to a set of nodes. Is there some way to do this? The only other thing I could think of was perhaps periodically running some cluster re-route script to try to keep the primary shards on the SSD machines but that seems pretty hacky and I'm not sure it would actually work reliably.
-- You received this message because you are subscribed to the Google Groups "elasticsearch" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To view this discussion on the web visit https://groups.google.com/d/msgid/elasticsearch/2d921ec3-8e0f-4a6f-aa7f-80e8dd7896e3%40googlegroups.com. For more options, visit https://groups.google.com/d/optout.
