We are currently sending a synchronous copy of deduplicated data to Amazon S3 and have a lifecycle policy to move the data to glacier after 1 day and the retention is set to 90 days for that secondary copy. I understand the need to seal the dedup db every now and again because glacier items cannot be pruned. What I am not understanding is how those blocks of data on the glacier side end up getting removed once you seal the dedup db.
Also, thinking ahead, is there an easier way to clean this stuff up in glacier? Maybe setup a bucket to manually clean/expire data in glacier (Amazon side) every 100 days and enable the Do not Deduplicate against objects older than 90 days? Would the pruning of the local AWS DDB happen still? Does anyone have any success at keeping only a DR copy in Amazon using LCP?