That last migration process is probably working on the data for a single nodename, where there can be no more than one migration process for one nodename.
It may be that there is still a lot of new data entering that disk pool such that migration is still going on; or, the tape/drive has issues which is making writing to it difficult and prolonged. Being write-shareable, disk pools involve a lot of locking, which adds overhead, as does caching, which can further prolong runtimes. Richard Sims
