We should make this consistent:
1) Either exclude T0_.Disk from the source nodes in Moves in RequestAllocator, so it is skipped
2) Or enable deletion of T0._Disk replicas with Moves in BlockDelete
I vote for 1 - T0_.*_Disk replicas are special (they are the first replica of collision data) and should be protected; and anyway the Tier0 is actively managing deletions of data from T0 Disk
There's currently an inconsistency between the policy and the deletion logic for Move requests from T0_.*_Disk nodes.
The Request logic allows T0_.*_Disk as source for the Moves, and will remove the source subscription from T0 Disk when the Move Request is approved:
https://github.com/dmwm/PHEDEX/blob/master/perl_lib/PHEDEX/RequestAllocator/Core.pm#L412
However the BlockDelete agent won't actually execute the deletions, leaving the data unsubscribed on T0_.*_Disk:
https://github.com/dmwm/PHEDEX/blob/master/perl_lib/PHEDEX/BlockDelete/Agent.pm#L74
We should make this consistent: 1) Either exclude T0_.Disk from the source nodes in Moves in RequestAllocator, so it is skipped 2) Or enable deletion of T0._Disk replicas with Moves in BlockDelete
I vote for 1 - T0_.*_Disk replicas are special (they are the first replica of collision data) and should be protected; and anyway the Tier0 is actively managing deletions of data from T0 Disk