I am looking to setup a small cluster of couchbase servers for a pet project that requires quite some storage that will be rarely accessed. So I want this data stored on spinning disks while I want my actual used data stored on SSD. (think that the HDD data is like 20x bigger than the data stored on SSD). So ideally I want 2 buckets but then on separate locations.
I couldn’t find anything specific on how to do this except this old article:
I could always hack it to mount the HDD into the specific data folder of the ‘cold storage’ bucket but not sure if that is the right way to go.
I think addressing this with filesystem mounts should be just fine. That’s probably the best bet.
Personally, I use ZFS with an SSD as a cache device. This is a Linux VM on that OS, but there are other ZFS options. That gives me the kind of fast access automatically. Couchbase will still have linear-scan type operations, but the adaptive replacement cache it uses will avoid caching these pages on the SSD.