On Wed, Jul 16, 2014 at 4:45 PM, Craig Lewis <clewis at centraldesktop.com> wrote: > One of the things I've learned is that many small changes to the cluster are > better than one large change. Adding 20% more OSDs? Don't add them all at > once, trickle them in over time. Increasing pg_num & pgp_num from 128 to > 1024? Go in steps, not one leap. > > I try to avoid operations that will touch more than 20% of the disks > simultaneously. When I had journals on HDD, I tried to avoid going over 10% > of the disks. > > > Is there a way to execute `ceph osd crush tunables optimal` in a way that > takes smaller steps? Unfortunately not; the crush tunables are changes to the core placement algorithms at work. If I blue-sky it I suppose we could do something like ship multiple CRUSH configurations and the hash ranges within which you use each one, and then incrementally move the dividing lines from 0 over to 1...but that'd be a big change to get protocol support (*none* of the currently-deployed software could work with a cluster in such a hypothetical state) and test everything with it, and is not on anybody's roadmap. I wouldn't expect anybody to be changing CRUSH tunables on a non-toy cluster; there's a reason we had the config option for whether you get warnings or not. :) -Greg Software Engineer #42 @ http://inktank.com | http://ceph.com
 |