CFA workflow for HPC -> JASMIN disk -> JASMIN tape -> JASMIN disk

[bnlawrence]

We need to document the expected workflow for data as

• a cycle arrives from HPC onto JASMIN transfer disk
• a CFA file is written for that data
• a new CFA file is written which includes existing cycles (probably on tape) and the new data
• user facing CFA files are updated per frequency
• the data is migrated to tape
• some data is brought back from a tape
• a CFA file is updated for the presence of a new Quark (which CFA file)

A sketch of this is here:

The initial use case is CANARI, but we want this to be a generic workflow.

[bnlawrence]

Here are a couple of flowcharts expressing some of the stuff on that white board

[bnlawrence]

Already seen an error. Fragments are not moved from tape to disk, they're copied. Hence the stuff on disk in GWS is always a replicant of stuff on tape. Will update these and stick them in the repository.