On Tue, Feb 19, 2013 at 12:55 PM, Dimitri Maziuk <dmaziuk@xxxxxxxxxxxxx> wrote: > On 02/19/2013 02:49 PM, Gregory Farnum wrote: >> On Tue, Feb 19, 2013 at 12:43 PM, Dimitri Maziuk <dmaziuk@xxxxxxxxxxxxx> wrote: >>> On 02/19/2013 02:39 PM, Sam Lang wrote: >>>> On Tue, Feb 19, 2013 at 2:31 PM, Dimitri Maziuk <dmaziuk@xxxxxxxxxxxxx> wrote: >>>>> PPS: >>>>> >>>>> on the node that has osd.0: >>>>> --------------------------- >>>>> >>>>> # cephfs /ceph/blastdb set_layout -u 2147483648 -s 2147483648 -c 2147483648 >>>> >>>> Try 2147483647. >>> >>> # cephfs /ceph/blastdb set_layout -u 2147483647 -s 2147483647 -c 2147483647 >>> Error setting layout: Invalid argument >> >> cephfs has a crappy interface, (which will shortly be obsoleted thanks >> to the new "virtual xattrs" which expose the layout stuff through >> xattrs). You need to specify the pool ID as well. :( > > Thanks but no, that's not it and besides it worked with 2147483648. > Looks like it's because it's not a multiple of some power of 2. Sorry, I hadn't read the whole story, just saw that last message. I'm not sure where it's coming from exactly, but honestly I wouldn't recommend running with objects that large. I'd expect you to see better performance (even over the network) with a bit more of a distribution. -Greg _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com