On Fri, Feb 19, 2016 at 5:11 AM, Steffen Weißgerber <weissgerb...@ksnb.de> wrote: > > >>>> Gregory Farnum <gfar...@redhat.com> schrieb am Montag, 8. Februar 2016 um > 19:10: >> On Mon, Feb 8, 2016 at 10:00 AM, Dzianis Kahanovich >> <maha...@bspu.unibel.by> wrote: >>> I want to know about plain (not systemd, no deployment tools, only own >> simple >>> "start-stop-daemon" scripts under Gentoo) upgrade hammer to infernalis and >>> see >>> no recommendations. Can I simple node-by-node mon+mds+osd's restart or need >> some >>> strict behaviour global per-service restart? >>> >>> PS "setuser match path = /var/lib/ceph/$type/$cluster-$id" added to config. >> >> All the upstream testing in order upgrades the monitors, then the >> OSDs, then RGW/clients/MDSes. Other than that you should be good just >> restarting processes, yes. >> -Greg >> _______________________________________________ >> ceph-users mailing list >> ceph-users@lists.ceph.com >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > Hello, > > when updating the clients after the cluster nodes would that increase the > probability for issues regarding missing features so that clients with older > ceph versions will be disconnected by the updated cluster? > > Are new features activated automatically during a cluster update?
Anything that might not be compatible with clients requires an administrator command to activate. -Greg _______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
