Failed to restart cluster with old "untouched" tablespaces

[diegosalvi]

Failed to restart cluster with old "untouched" tablespaces

[diegosalvi]

I think we should avoid this issue at all... It make no sense to attempt a recovery from a very old position of time when really there is nothing to recover.
We should "advance" log position on metadata to a recent position every time we make a checkpoint. With no new data we should try to advance to next "current" position (just we must be sure to not skip a concurrent incoming update/insert/delete)

[eolivelli]

I cannot understand this problem, it should not happen by design and we should have test cases.
which is the error on the bookie ? did you drop some bookie without "decommissioning" it ?

we are not deleting the "active ledgers", so the Leader is always able to recover.
If a follower has an old snapshot then it downloads a snapshot from the Leader.