Hi all,
We recently reviewed our storage consumption and one thing I realized is that, apart from running on out-of-warranty, the nfs share used by cbs.centos.org koji is growing (which is a good sign that people/SIGs are using it) but also with things that shouldn't stay behind, like temporary scratch (test) builds. Problem is that apart from consuming disk space on that nfs storage node, it's also consuming resources on the (older) backup node, also extending the needed time to take backup.
I intend (in the following days) to implement through ansible a script itself just based on upstream koji-prune-scratch one (https://pagure.io/koji-tools/blob/master/f/src/bin/koji-prune-scratch)
I'd like to start cleaning (deleting) scratch jobs that are (like in that upstream script) more than 7 days (scratch builds aren't proper builds and so are never tagged and so never signed/pushed out to mirrors).