transition between pre-production CephFS to production CephFS

[ickc]

From a meeting between SO:UK DC and Blackett people today:

• Blackett is currently using DPM and will transition into Ceph in 2024.
• Currently, XRootD is serving from DPM, which will transition into serving from Ceph instead.
• Ceph exposes the filesystem via CephFS (currently xrootdfs can be used to mount XRootD as FS via FUSE, but not on worker nodes), available on all nodes including worker nodes.
• pre-production CephFS will be put on our machines (the SO:UK DC purchased machines), but may be subject to purging when transitioning into production
  • This will not be a problem for us (stopping us using pre-production CephFS ASAP), and we plan to copy data for our users. Both can be mounted, and then say we make the pre-production CephFS read-only, and replicates it to the production CephFS. We should provide env. var. to hide the differences in absolute path (see #23)

Edit: relevant context: I find this when surfing the net, probably relevant: Ceph Deployment and Monitoring at Lancaster, GridP47, 23 March 2022, Gerard Hand, Steven Simpson, Matt Doidge. I like this bit:

There’s only so many times you can declare a million files lost and not rethink all your life choices.

[ickc]

Closing. Might reopen after we have a pre-production storage solution.