AFAIR, having multiple public or cluster Ceph networks will not magically provide switch failover. Multiple netwworks will merely tell Ceph daemons where to listen to and which networks will be used to talk both with Ceph daemons and clients, no more, no less. In fact, you will need to route both networks so a daemon can talk to every other Ceph component on either two networks.
To accomplish switch redundacy you must use link aggregation / bonding [1] (ideally LACP if your switch support stacking), then use a single public network and (optionally but recomended) a single cluster network.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.