Owing to budgetary issues and the vagaries of delivery schedules, we have decided to clear off as much of our existing disk space as we can, up to ~100 TB.
Some 40 TB is taken up by I&T and related datasets. Due to the interaction with the eLog database, we will address this space last. We'll target MC tasks first. We don't the manpower or time to copy all deleted data to tape. Once something is gone, it is gone! Note that we will need to be reasonably aggressive on deleting old tasks since it is the more recent ones that take up the bulk of the space.
We will try to lump datasets into three categories: nuke in toto; keep merit tuples; keep fully. For the "keeps" we would move the datasets into xrootd. "Keep fully" is for recent datasets still in active use; "keep merit" is for key datasets we may wish to refer to later.
The opening algorithm for MC is to keep the most recent of the standard tasks: allGamma, allMuon, sampled day, backgrounds. We can add a proviso that tasks created this February will be kept. We should also keep key datasets, eg those used to make the Pass4 analysis & irfs.
Standards
allGamma: allGamma-GR-v13r9p3-Lyon, allGamma18GeV-GR-v13r9p3-Lyon
allMuon: allMuon-GR-v13r9
sampled Day: backgnd-GR-v13r9-Day
backgrounds: backgnd-GR-v13r9
Major DataSets
OpsSim2/Big Run backgrounds: opssim2-GR-v13r9p3
GRB grid: GRBgrid-GR-v13r9p4-Lyon