Dear colleagues, we build L3 topology for use with CEPH, which is based on OSPF routing between Loopbacks, in order to get reliable and ECMPed topology, like this: 10.10.200.6 proto bird metric 64 nexthop via 10.10.15.3 dev enp97s0f1 weight 1 nexthop via 10.10.25.3 dev enp19s0f0 weight 1 where 10.10.200.x are loopbacks (plenty of on every node) and 10.10.15.x/25.x are interface addresses of the host. So, my physical interface addresses on this host are 10.10.25.2 and 10.10.15.2, local loopback is 10.10.200.5. CEPH configured in the way [global] public_network = 10.10.200.0/24 [osd.0] public bind addr = 10.10.200.5 cluster bind addr = 10.10.200.5 but regardless of these settings ceph-osd process originates connections from interface addresses, e.g. tcp 0 0 10.10.15.2:57476 10.10.200.7:6817 ESTABLISHED 52896/ceph-osd tcp 0 0 10.10.25.2:42650 10.10.200.9:6814 ESTABLISHED 52896/ceph-osd tcp 0 0 10.10.25.2:36422 10.10.200.7:6804 ESTABLISHED 52896/ceph-osd tcp 0 0 10.10.15.2:49940 10.10.200.6:6815 ESTABLISHED 52896/ceph-osd which has a negative impact when one or another physical port gone. Is there way to say CEPH to originate connections from specified ip address, which is unbinded from physical infra? Thank you. _______________________________________________ ceph-users mailing list ceph-users@xxxxxxxxxxxxxx http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
 |