Ceph

To assist with CephFS disaster recovery, provide the ability to dump an entire pool (the cephfs metadata pool) to a file, and subsequently restore that to an empty pool. This should include all attrs and omaps as well as object data.

This should come with a test that stops a running filesystem, dumps the pool, wipes the pool, restores it, and checks that filesystem comes up again cleanly as if nothing happened.

I expect some issuing of requests in parallel will be needed here to get a sensible level of throughput.

Clearly this is only practical for pools containing somewhat small amounts of metadata: should probably try to make it clear in any docs that this is not meant to be a general purpose way of backing things up!

Possible extension: optionally stream backup to another RADOS pool (or another ceph cluster) instead of a file.