There was this discussion once that as you get enough objects in your cluster with an delete-at on them you'd rather just catch it on the walk then bother with the indexing and object-expirer daemon. Having delete-at cleanup baked into the auditor seems like a safe and prudent stride.
One thing missing there is the object-auditors don't really have any guarantee of finishing because they randomly select paths every time they start. I've been wanting to fix bug # 1183656 - I can commit to that this cycle.
+1 to doing in the auditor
There was this discussion once that as you get enough objects in your cluster with an delete-at on them you'd rather just catch it on the walk then bother with the indexing and object-expirer daemon. Having delete-at cleanup baked into the auditor seems like a safe and prudent stride.
One thing missing there is the object-auditors don't really have any guarantee of finishing because they randomly select paths every time they start. I've been wanting to fix bug # 1183656 - I can commit to that this cycle.