Basically the title says it all - how many replicas do you use for your storage pools? I've been thinking 3 replicas for vms that I really need to be confident of the data durability due to hardware issues and 2 replicas for where I need speed but less redundancy. I was wondering what other people have chosen and why?
Ceph - how many replicas do you use?
- Thread starter nethfel
- Start date
I would agree, 2 similar to raid1, 3 equal better durabiliity, all depends on how much concurrent failures you would like to be able to handle while keeping service available, price/durability vs price/perf I guess 
I think that more replicas does not mean less speed. Read requests are spread between all replicas, so more replicas -- higher read speed. But you have more network load on write.
So it's more tradeoff between durability and price.
So it's more tradeoff between durability and price.
Hi,
AFAIK still not yet - the reads are from the primary only. And write speed drops with each replica, because of one more writing before ACK.
To answer the thread question, I started with replica 2 but with many OSDs it's a little bit dangerous, because 2 failed disks mean data loss (and not only some VM, because the data of all VMs nearly on all OSDs).
Now I use replica 3 and one EC-pool for archive-data with m=2, which mean that also 2 ODSs can fail without data loss.
Udo