Hi Joakim, On Mon, Dec 5, 2016 at 1:35 PM <joa...@verona.se> wrote:
> Hello, > > I have a question regarding if Ceph is suitable for small scale > deployments. > > Lets say I have two machines, connected with gbit lan. > > I want to share data between them, like an ordinary NFS > share, but with Ceph instead. > > My idea is that with Ceph I would have redundancy with two machines > having complete copies of the data. I also imagine that the > performance could be quite okay in principle, depending on how Ceph > works, which I'm not quite sure of. > > A use case would be to share my home directory on two machines, or > three maybe. > > A workload I'm concerned with is software builds. Would Ceph be > competetive in this use-case as compared with a local disk? As far as > I can tell Ceph doesn't use the "eventual consistency" approach. Does > that mean that all writes have to sync over all the nodes in the Ceph > cluster, before the write can be considered complete? Or is one node > enough? > > /Joakim > > _______________________________________________ > ceph-users mailing list > ceph-users@lists.ceph.com > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > I have done such setups with one high quality machine. If stability is a concern, without an absolute requirement for multi head node, simple is a really good approach. You would have to assess downtime requirements, but I can tell you that we have over 30 24/7 systems out there running not much more that NFS on ZFS, and these are great workhorses, with little or no downtime. HA features to set up - good RAID controller (we like Areca), bonding, ideally with 10GbE or more - active/backup or LACP. ECC RAM, and monitor the hardware. Take good backups. I have a decent lab Ceph setup with 3 OSD nodes and virtual machines as MONs, but running off different storage. I would not do Ceph with 2 OSD nodes. As Christian replied, DRBD has been widely used for two node setup. Good luck! Alex
_______________________________________________ ceph-users mailing list ceph-users@lists.ceph.com http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
