Sharded Cluster Snapshot Restores - Throttling and Src/Dst Mapping
When restoring a sharded cluster snapshot, provide a means of mapping the source shard replica set names to the target shard replica set names. This will allow users to predictably restore large/small shards to the appropriate target hosts.
Currently, this can be done for sharded clusters with fewer than 10 shards by naming the shards in a predicable shard## pattern. However, greater than ~10 shards leads to an alphabetical => alphanumeric restore plan (e.g. shardA10 restores to shardB2).
Restoring large sharded clusters can also overwhelm networks where MongoDB Agents are downloading snapshots from Ops Manager(s) at the same time. Please consider providing a mechanism to throttle the number of Agents concurrently downloading snapshots with the understanding that this will come at a cost of longer restore times. This is helpful for large restores scheduled on a regular basis with expected downtime of a QA or UAT environment where uptime is not a priority.