> We've got some QEMU/KVM work load linked to librbd 14.2.21 (with > backported patches), being complicated to "just upgrade" for the time being. > > With regard to compatibility, I'm curious about cluster versions. Are > there specific features or other caveats to look out for when connecting > such work load to an Octopus, a Pacific, a Quincy or even a Reef cluster?
My guess is that it probably never would be a problem to run a newer client for an RBD image made on an older cluster, since they do set feature flags on the images, and an old release would only set flags it know about, and the later versions would of course handle them, but the opposite might not be equal, so it a Reef client creates an RBD image it might ask for flags that a later Oct client doesn't understand what to do with. And as far as ceph stuff goes, RBD is (when used with mostly defaults) is quite simple and the protocol or stored data isn't changing much between releases. It does develop new features like being able to be stored on EC pools, fast diffs and things like that, but the basic read and write operations are quite the same. -- May the most significant bit of your life be positive. _______________________________________________ ceph-users mailing list -- [email protected] To unsubscribe send an email to [email protected]
