That should be good. The two main things to think about are:

1. etcd stores keys in RAM (and snapshots to disk), so you need enough of that 
for your data set
2. etcd doesn’t serve a successful write until it is written to disk on a 
quorum of machines, which means it is IOP-sensitive. Larger instances generally 
get more IOPs. SSDs will help here too.

 - Rob

> On Aug 24, 2016, at 7:08 AM, Norman Khine <[email protected]> wrote:
> 
> i am trying to set a production 5 node coreos cluster and wanted to know what 
> is a recommended instance type that i should use? currently i am using 
> c4.large
> 
> any advise much appreciated

Reply via email to