It was recommended to set sort_bitwise in the upgrade from Hammer to Jewel when Jewel was first released. 10.2.6 is definitely safe to enable it.
On Tue, Jul 18, 2017, 8:05 AM Dan van der Ster <d...@vanderster.com> wrote: > Hi Martin, > > We had sortbitwise set on other jewel clusters well before 10.2.9 was out. > 10.2.8 added the warning if it is not set, but the flag should be safe > in 10.2.6. > > -- Dan > > > > On Tue, Jul 18, 2017 at 11:43 AM, Martin Palma <mar...@palma.bz> wrote: > > Can the "sortbitwise" also be set if we have a cluster running OSDs on > > 10.2.6 and some OSDs on 10.2.9? Or should we wait that all OSDs are on > > 10.2.9? > > > > Monitor nodes are already on 10.2.9. > > > > Best, > > Martin > > > > On Fri, Jul 14, 2017 at 1:16 PM, Dan van der Ster <d...@vanderster.com> > wrote: > >> On Mon, Jul 10, 2017 at 5:06 PM, Sage Weil <s...@newdream.net> wrote: > >>> On Mon, 10 Jul 2017, Luis Periquito wrote: > >>>> Hi Dan, > >>>> > >>>> I've enabled it in a couple of big-ish clusters and had the same > >>>> experience - a few seconds disruption caused by a peering process > >>>> being triggered, like any other crushmap update does. Can't remember > >>>> if it triggered data movement, but I have a feeling it did... > >>> > >>> That's consistent with what one should expect. > >>> > >>> The flag triggers a new peering interval, which means the PGs will > peer, > >>> but there is no change in the mapping or data layout or anything else. > >>> The only thing that is potentially scary here is that *every* PG will > >>> repeer at the same time. > >> > >> Thanks Sage & Luis. I confirm that setting sortbitwise on a large > >> cluster is basically a non-event... nothing to worry about. > >> > >> (Btw, we just upgraded our biggest prod clusters to jewel -- that also > >> went totally smooth!) > >> > >> -- Dan > >> > >>> sage > >>> > >>> > >>>> > >>>> > >>>> > >>>> On Mon, Jul 10, 2017 at 3:17 PM, Dan van der Ster <d...@vanderster.com> > wrote: > >>>> > Hi all, > >>>> > > >>>> > With 10.2.8, ceph will now warn if you didn't yet set sortbitwise. > >>>> > > >>>> > I just updated a test cluster, saw that warning, then did the > necessary > >>>> > ceph osd set sortbitwise > >>>> > > >>>> > I noticed a short re-peering which took around 10s on this small > >>>> > cluster with very little data. > >>>> > > >>>> > Has anyone done this already on a large cluster with lots of > objects? > >>>> > It would be nice to hear that it isn't disruptive before running it > on > >>>> > our big production instances. > >>>> > > >>>> > Cheers, Dan > >>>> > _______________________________________________ > >>>> > ceph-users mailing list > >>>> > ceph-users@lists.ceph.com > >>>> > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > >>>> _______________________________________________ > >>>> ceph-users mailing list > >>>> ceph-users@lists.ceph.com > >>>> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > >>>> > >>>> > >>> > >> _______________________________________________ > >> ceph-users mailing list > >> ceph-users@lists.ceph.com > >> http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
