On Sat, 9 May 2015, Kyle Bader wrote: > It looks like the NUMA problem [1] could easily be solved by: > > #include <numa.h> in ceph-osd.cc > numa_available() should return 0 > read from /var/lib/ceph/ceph-$id/numa_node > make sure int > make sure sure numa node exists > int numa_bind(numa_node) > > Then that OSD will only use the cpu cores and memory in that NUMA > region. This could also be handy for converging ceph-mon/rgw/etc with > OSD. > > [mon.a] > numa_node = 1 > > ceph-mon numa_bind(1) > > [rgw.a] > numa_node = 1 > > radosgw numa_bind(1) Hmm! So, numa_node default would probably be -1. What about a zone_reclaim_mode option? Or should we just do a warning? sage > > [1] https://www.mail-archive.com/ceph-users@xxxxxxxxxxxxxx/msg16274.html > > -- > Kyle Bader - Red Hat > Senior Solution Architect > Ceph Storage Architectures > > -- To unsubscribe from this list: send the line "unsubscribe ceph-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
