On Fri, 15 Sep 2017, Two Spirit wrote: > Sage, My email is bounced back saying there is a virus. I don't think the > mailing list likes ascii art so I'm resending directly to you. > cephfs with posix fs works great > > [ host1 auto/mount ] --+-- [ceph 1] > | > [ host2 auto/mount ] --+-- [ceph 2] > | > +-- [ceph 3] > > automount can automatically connect to available ceph node. Network device > that doesn't support ceph or cephx needs access to ceph, I can export it > through NFS/SMB > > NetAppliance ----exports--[ host1 auto/mount ] --+-- [ceph 1] > | > --[ host2 auto/mount ] --+-- [ceph 2] > | > +-- [ceph 3] > > > host 1 becomes a single point of failure. host 1 goes down, > ceph=unavailable. Solution is to create virtual ip with heartbeats for high > availability. > > NetAppliance --[VIP]--+---[ host1 auto/mount ] --+-- [ceph 1] > | | > +---[ host2 auto/mount ] --+-- [ceph 2] > | > +-- [ceph 3] > > Is this scenario possible with CephFS since thet NetAppliance doesn't > support mounting ceph nor cephx, nor round robin mounting multiple IP. > > [ host1 auto/mount ] --+-- [ceph 1] > | > [ host2 auto/mount ] --+-- [ceph 2] > | > [ NetAppliance ] ----+-- [ceph 3] NetAppliance == host re-exporting cephfs via nfs. The recommended way to do this is with nfs-ganesha. There is a bunch of info about how to do this in an HA way (with N active servers even) but I'm not sure it is well collated on docs.ceph.com. sage
