Hi Jordi, I would like to propose you consider use any virtualization layer on top of ceph's RBD cluster. So you will be able achieve huge high availability during planned and unplanned downtime.
And as mentioned Gregory already, RBD have ability to take snapshots. You can easily export crush-consistent snap of running VMs and import they when would be necessary back to the complex. BTW, we testing right now ability to take consistent backups via guest agents behavior. On Fri, Mar 21, 2014 at 11:22 PM, Gregory Farnum <[email protected]> wrote: > When starting this you should be aware that the filesystem is not yet > fully supported. > > > On Thursday, March 20, 2014, Jordi Sion <[email protected]> wrote: > >> Hello, >> >> I plan to setup a Ceph cluster for a small size hosting company. The aim >> is to have customers data (website and mail folders) in a distributed >> cluster. Then to setup different servers like web, smtp, pop and imap, >> accessing the the cluster data. >> >> The goals are: >> >> * Store all data replicated across different nodes >> * Have all data accessible for every server (like www servers). This way, >> we can easily move a web from a server into another, or from, let's say >> apache to nginx. Or have all email accounts accessible from every pop/imap >> server. >> >> I am about to build a 3 node cluster to start tests: 1 MDS with 240Gb SSD >> and 2 OSD+Monitor with 2x2Tb, with 32 Gb of Ram each and will be >> interconnected with a 1Gb Private LAN. >> > > The MDS doesn't need any local storage beyond a few config files. :) > > >> >> Mainly, the servers using the cluster will provide Web serving, FTP >> access and Email SMTP, POP and IMAP. Also I need to provide MySQL database, >> which I am not sure how data fits in a Ceph cluster. >> >> I have some questions: >> >> 1) The plan is to keep MDS node dedicated. Will OSD's be able to act as >> webservers (apache and proftpd) or mail servers (postfix, dovecot, amavis >> and spamassassin). >> > > That will depend on how much CPU they have and what clients you're using > (you don't want to loop back mount with a kernel client ). > > >> 2) How can I manage to have MySQL data stored in Ceph? Is that a good >> idea? Any suggestions? >> > > I'd recommend just using RBD rather than CephFS. That'll give you a block > device which you can mount anywhere (but only on one at a time). > > > >> 3) To prevent major disasters, What is a good practice/strategy to >> backup/replicate data in the cluster? >> > > Hmm, there's not a good tailored answer for CephFS. With RBD there are > some options around snapshots and incremental diffs. > -Greg > > >> >> Thanks in advance, >> Jordi >> > > > -- > Software Engineer #42 @ http://inktank.com | http://ceph.com > > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com > > -- Igor Laskovy facebook.com/igor.laskovy studiogrizzly.com
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
