Local patch access to CEPH from PVE Node possible?

[DC-CA1]

In scenario where 3 PVE host have CEPH running and each of them have Direct attached disks , trough SAS for example ,

will the PVE Node benefit from been able to access the DATA localy from the HBA controller and not passing trough the network to loopback ?

we try to figure if we can Avoid local VMS to cross the Network for nothing when there a active copy available from the Same PVE node

 

[gurubert]

When your VM image is an RBD or even qcow2 on CephFS its data is spread over many many RADOS objects and stored on every OSD. In a Ceph cluster there is no local data.

 

[DC-CA1]

When your VM image is an RBD or even qcow2 on CephFS its data is spread over many many RADOS objects and stored on every OSD. In a Ceph cluster there is no local data.

Each of our nodes have Direct attached SAS and we where hopping that the Locals VM would not load the 10GB interface to access back the data. but as its spread over many node like RAID5 for example is than i understand its the only option.

it would be amazing if we could Sync our active datastore into another cluster but for now the only way seem to do ZFS storage based but it make no sense for us to do that for large VM with a 16 node cluster that will need to have a copy of that VM on each .
been stuck to use ZFS only to enable remplication is really sad ,.